FTL REPORT R82-3
FTL COPY, DON'T REMOVE
33412, MIT
DECISION SUPPORT SYSTEMS FOR
AUTOMATED TERMINAL AREA
AIR TRAFFIC CONTROL
JOHN DEMETRIOS PARARAS
SEPTEMBER 1982
....
02139
FTL REPORT R82-3
DECISION SUPPORT SYSTEMS FOR AUTOMATED TERMINAL AREA
AIR TRAFFIC CONTROL
John Demetrios Pararas
September 1982
-
blank page -
-2-
ABSTRACT
This work studies the automation of the terminal area Air Traffic
The ATM/C decision-making
Management and Control (ATM/C) system.
a number of "automation
into
down
broken
and
process is analyzed
with particular
described
is
functions
these
Each of
functions".
with the
interactions
its
and
system
overall
the
emphasis on its role in
Flight
Traffic
and
Runway Scheduling
other ATM/C automation functions.
greatest
the
with
functions
two
the
as
Plan Generation are identified
improvements over the current
potential for providing efficiency
terminal area ATC system and are studied in detail.
A very general Mixed Integer Linear Programming (MILP) formulation
Less general
is developed.
Scheduling problem
the Runway
of
formulations and algorithms which have appeared in the literature are
A heuristic algorithm is developed. The
reviewed and evaluated.
of Dear and adopts tne Maximum Position
work
the
on
based
algorithm is
It extends Dear's work
him [DEA 76).
by
proposed
Shifting methodology
multiple runway
to
applicable
is
(1) it
ways:
several
in
simulation
real-time
a
in
operate
to
configurations. (2) it is designed
constraints
arbitrary
accept
to
environment, and (3) it is designed
imposed by the ATM/C controller.
The methodology for generating flight plans is developed. Flight
plans are based on a specified route structure. They are 4-dimensional
and conflict-free. To allow maximum runway scheduling flexibility, a
It is designed to allow easy
.specific route structure is proposed.
modification of flight plans to adapt to the dynamically changing
schedule.
To allow algorithmic development and testing of this (as well as
other) ATM/C automation concepts, a real-time terminal area simulation
The facility has a
facility (called TASIM) is designed and implemented.
number of characteristics which make it a good general purpose tool for
terminal area ATM/C research:
(1) Highly modular design which allows addition, removal and
modification of functions with relative ease.
(2) Realistic modelling of the aircraft dynamics of motion
and the aircraft guidance system. Errors introduced by
the navigation equipment (onboard and on the ground) and
by the surveillance radars are also modelled.
-3(3) Capability to simulate multiple controller positions
(4) Flexible controller interface which allows easy implementation
of alternative displays and alternative protocols for manmachine interaction.
The simulation is fully operational in the conventional (manual)
with an
it is
currently interfaced
In addition,
mode.
ATC
a
special
and
with
implementation of the runway scheduling heuristic,
purpose final vectoring display designed to aid the controller in
precisely timing the delivery of landing aircraft at the outer marker.
-4-
Acknowledgement
source of advice and
Robert Simpson has been a
Professor
He has been the teacher
work.
this
of
course
the
throughout
inspiration
Control, and patiently
Traffic
Air
of
field
the
to
me
who introduced
the last few years. He
over
subject
the
on
relayed his vast knowledge
most needed it, and at
I
when
guidance
has been the advisor who provided
my own ideas. Working
pursue
to
freedom
the same time allowed me the
and pleasing
rewarding
under his supervision has been an extremely
grateful.
experience. For this I am deeply
Professor Amedeo Odoni has shaped my entire graduate career at MIT.
My admiration for his rare combination of academic achievement, teaching
ability, and deep love and respect for people, was immediate and has
kept growing over the years. I feel privileged to have known him and
to be able to call him my friend.
Professor Thomas Magnanti sparked and has kept alive my interest in
mathematical programming and optimization. His inspired teaching and
vaiuable guidance during my years at MIT are deeply appreciated.
I would also like to express my thanks to Gabriel Handler who
introduced me to the field of flight transportation and is thus
indirectly responsible for bringing me among the outstanding group of
people which comprise the Flight Transportation Laboratory at MIT.
Finally, my wife Jan and my son Dimitri provided the stability
which saw me through the difficult moments and multiplied my feeling of
accomplishment during my successes.
The research for this project was performed at the MIT Flight
was
Funding for the development effort
Transportation Laboratory.
University
Joint
the
through
Center
provided by NASA Langley Research
Program for Air Transportation Research. (NGR-36-009-017)
-
blank page -
406MMAN009W
-5-
TABLE OF CONTENTS
Page
CHAPTER 1:
INTRODUCTION
14
1.1
Background and Motivation
14
1.2
Overview of Previous Research
19
1.3
Document Summary
25
CHAPTER 2:
FUNCTIONAL DESCRIPTION OF AN AUTOMATED TERMINAL AREA
ATM/C SYSTEM
29
2.1
Introduction
29
2.2
The Generic Terminal Area ATM/C System
30
2.3
The ATM/C Flight Plan Generator
38
2.3.1
Nominal Flight Plan Generator
40
2.3.2
Runway Scheduler
41
2.3.3
Traffic Flight Plan Generator
44
2.4
ATM/C Command Processor
45
2.5
The ATM/C Traffic Monitor
49
2.5.1
Conformance Detection and Resolution
49
2.5.2
Hazard Detection
50
-6-
TABLE OF CONTENTS
(continuea)
Page
CHAPTER 2
(continued)
Hazard Resolution Path Generator
2.5.3
52
2.6
Traffic and CTID Displays
53
2.7
The ATM/C Controllerts Role
55
CHAPTER 3:
A SIMULATION TESTBED FOR THE DEVELOPMENT
OF ATM/C AUTOMATION SOFTWARE
57
3.1
Overview
57
3.2
The Hardware Environment
58
3.3
Major System Components and Interfaces
60
3.4
The Simulation Monitor Process
64
3.4.1
Data Base Initialization
64
3.4.2
Interprocess Communication
68
3.4.3
Subordinate Processes and Execution Control
70
3.4.4
Initialization or Terminal I/0
73
3.4.5
Timer Loop Alert
74
3.4.6
Delay Timer Alert
76
3.4.7
Screen Update Timer Alert
76
3.4.8
Mailbox Message Processor
78
-7TABLE OF CONTENTS
(continued)
-Page
CHAPTER 3
(continued)
Terminal Input Processing
79
The Position Generator Process
81
3.4.9
3.5
3.5.1
Mailbox Message Processor
82
3.5.1
Traffic Generation Model
83
3.5.3
Aircraft Status Update and Handotf Processing
86
3.5.4
Aircraft Dynamics
87
3.5.5
Air Data System Update
88
3.5.6
Aircraft Navigation System
88
3.5.7
Surveillance System
89
The ATM/C Controller Process
91
3.6.1
Mailbox Message Processor
93
3.6.2
Command Activation
94
3.6.3
Traffic Display Driver
97
3.6.4
Input Editor
102
3.6.5
Input Processor
102
3.6.6
CTID Update
104
3.6
3.7
Implementation Status
106
-
-
W-
-8-
TABLE OF CONTENTS
(continued)
Page
CHAPTER 4:
THE RUNWAY SCHEDULING PROBLEM
107
4.1
Introduction
107
4.2
Definitions
109
4.3
Formulation ot the Dynamic Runway Scheduling Problem
113
4.4
Measures of Runway Efficiency
117
4.4.1
Runway Capacity
11t
4.4.2
Aircraft Delays
120
4.4.3
User Cost Distribution: The CPS Methodology
125
4.5
The Complexity of the Runway Scheduling Problem
128
4.6
Variations ot the Runway Scheduling Problem
134
4.6.1
Scheduling Landings on a Single Runway (MPS=infinity)
135
4.6.2
Scheduling Landings on a Single Runway (MPS<Na-1)
138
4.6.3
Scheduling Landings on Independent Runways (MPS<Na-1)
143
4.7
Heuristic Versus Exact Algorithms
145
4.8
A Heuristic Algorithm for the Scheduling Problem
150
4.8.1
The Initialization Stage
1b1
-9TABLE OF CONTENTS
(continued)
Page
CHAPTER 4
(continued)
4.8.2
The Feasibility Stage
154
4.8.3
The Optimization Stage
156
4.9
Implementation Status
162
4.10
Coordination Among Terminal Area Control Sectors
163
4.11
Computer Aided Vectoring for Approach Spacing
166
CHAPTER 5:
A IETHODOLOGY FOR TRAFFIC FLIGHT PLAN
171
GENERATION
5.1
Introduction
171
5.2
Flight Path Structures
175
5.3
Ground Track Selection and Speed Control
104
5.4
Altitude Selection
188
5.5
Conflict Identification and Resolution
189
5.6
The Base and Final Approach Phases
192
5.7
Aircraft Guidance
207
5.7.1
Recovery from Errors in the Execution of' Turns
207
5.7.2
Speed Control
211
I
-10TABLE OF CONTENTS
(continued)
Page
CHAPTER 6:
CONCLUSIONS AND RECOMMENDATIONS
215
6.1
Summary
215
6.2
Recommendations for Future Research
219
APPENDICES
A
Interarrival Dynamics
222
B
The Kinematics of the Intercept Vector
229
REFERENCES
233
-11LIST OF FIGURES
Page
2-1
Overall Schematic Diagram of the Terminal Area ATM/C System
31
2-2
The Major Elements ot the Automated Ground Control System
35
2-3
Functional Block Diagram of the Automated Ground Control
System
39
3-1
Current TASIM Hardware Diagram
59
3-2
TASIM Software Processes
61
3-3
Top Level Flowchart of the Simulation Monitor Process
65
3-4
Timer Loop Alert Processing Detail
75
3-5
Delay Timer Alert Processing Detail
77
3-6
EXIT Message Processing Detail
80
3-7
Functional Diagram of the ATM/C Controller Process
92
3-8
Traffic Display Layout
99
3-9
Controller Tabular Information Display Layout
105
4-1
Runway Scheduling Heuristic (Initialization Stage)
152
4-2
Runway Scheduling Heuristic (Feasibility and Optimization
Stages)
157
4-3
Final Vectoring Display
168
5-1
Typical Terminal Control Area Layout
176
5-2
Typical Nominal Approach Flight Plan
177
-12-
LIST OF FIGURES
(continued)
Page
5-3
Example of an Approach Path Structure
182
5-4
Typical Flight Plan for the Base and Final Approach Phases
193
5-5
Horizontal Separations During the Final Approach Phase
197
5-6
The Kinematics of Aircraft Turns
208
5-7
Recovery from Early Speed Reduction in the Downwnind Phase
212
A-1
Minimum Interarrivai Separation (Overtaking Case)
225
A-2
Minimum Interarrival Separation (Opening Case)
225
B-1
The Kinematics of the Intercept Vector
230
-13LIST OF TABLES
Page
90
3-1
ATC Radar Beacon System Parameters
5-1
Effect of Aircraft Speed on Horizontal Separations and on the
Timing of Descent
202
A-1
Minimum Interarrival Separations at the Runway
226
A-2
A comparison of Landing Sequences
228
-
blank page -
-14CHAPTER 1
INTRODUCTION
1.1
Background and Motivation
Over the last twenty years advances in computer hardware have had a
tremendous impact on every aspect of life.
industry, due to its close ties
has traditionally
program,
practically every aspect of
assisted design and
with the defense industry and the space
been among the
technology.
advanced computer
The civil air transportation
frontrunners in
a crucial role in
Today, computers play
air transportation.
the use of
Starting from computer
all tne way to
manufacturing of aircraft and going
autopilots and automatic flight control systems, computers perform tasks
of increasing number and complexity.
Air
The
Traffic Control
experienced ,the same
automation
started
implementation
system.
impact.
over
of the
The
twenty
system in
years
TRACONs (Terminal Radar Control)
facilities respectively.
computers which
the United
effort towards
current (third)
Control Centers) are the current
flights.
(ATC)
ago
States has
air traffic control
with
generation Air
the
design
and
Traffic Control
and ARTCCs (Air Route Tratfic
ATC system's terminal area and enroute
Together they
provide a nationwide network of
store, maintain, and distribute
information on all IFR
-15In the present ATC system, automation has been accomplisned in data
and
collection
plan
Flight
processing.
automatically collected and distributed
and
data are
surveillance
Radar
to appropriate stations.
returns are automatically correlated and filtered so that the controller
Identity
has better position information than was previously available.
and altitude information, supplied by beacon radar for suitably equipped
aircraft, is associated with radar
returns.
All the available data are
used to generate clear alphanumeric traffic displays for the air traffic
controller.
flight
Generally,
productivity
air traffic
of the
controller by
Finally, it has
of many
In addition, it has
provided controllers with
up to
and clearances for
current and future flight plans
date information on
the
among ATC facilities thus allowing
enhanced the exchange of information
better coordination.
relieving him
up to then.
tasks which he had to perform manually
improved
has
automation
processing
IFR flights.
Further
capabilities
improvement
will be
Subsystem (ETABS).
of
current
the
achieved through
processing
data
system's
Tabular Display
the Electronic
The prototype software for ETABS is currently in its
late stages of development under a Federal Aviation Administration (FAA)
research
and
strips" that
designed
to
development
program.
are currently used
enhance
ETABS
will
replace
by the controller.
inter-controller communication
In
and
the "flight
addition it is
provide more
flexible data entry capabilities than are currently available.
-16It is generally accepted today that
is not sufficient
alone
controller's
the ATM/C system.
per
workload
terminal
area
limiting
factor determining
in
increases
aircraft
ATM/C system
the
at
Instead,
least, the
the system's
system's
capacity
will
that
human
today,
in the
element is not the
capacity.
be
capacity of the
it has decreased the
so
handled
This is
in data processing
any increase in the
date has not provided
in the
usually the bottleneck
Indeed the automation
entire ATM/C system.
facilities comprising
traffic.
required by increasing
the terminal area which is
particularly true in
achieved to
further improvements
provide the
the ATM/C system
capacity of
of the
to
automation in data processing
It follows tnat
achieved
only through
increased capacity of the airport and the ATC facilities up to the point
when the
capacity of these
controller.
Since
it is very
country's major airports
in capacity
elements becomes comparable to
have to be
unlikely that physical
will be feasible in the
the result of
that of the
expansion of the
future, the increases
more efficient use
of available
resources.
This
realization has
processing automation
decision-making for
resulted in
efforts to
and design computer
air traffic control.
go beyond
software which take
The most important
the data
part in
of these
efforts are reviewed in section 1.2.
(1) The recent air traffic controllers' strike and subsequent firing has
of course changed that situation but the effects are hoperully
transient.
-17-
which
to
some extent
from the
deviate
the following premises
motivated by
here was
work described
The
prevailing trends
in current
research on the subject of ATC automation:
traffic density in the
1. The high
terminal area gives high
priority to the problems of automating ATC decision-making
the
In addition since
a
for
responsible
system operation
that can be
there will offer much larger
an
the terminal area airspace is
of
percentage
any
air transportation,
in
experienced
large
of
should be much
for the enroute airspace
automated system
simpler.
development
the
airspace,
area
terminal
been found for
solutions have
Once satisfactory
there.
the
delays
improvements in
achieved through automation
benefits than those that can
be achieved through automation of the enroute ATC system.
should result in a
2. Decision automation
delays
spacing
reduction
aircraft
landing aircraft
of
of
delays by
during
optimal scheduling
greater
by aircraft.
experienced
potential
the
Even
true reduction of
though automated
will result
alone
avoiding excessive
final
approach
phase,
in some
gaps between
we believe
of all runway operations
to have much
reduction and
we therefore
for delay
consider it to be the primary ATC automation function.
We
propose, therefore, to view the problem not only as one or
-18-
stress this view we
order to
term
also one ot traffic
control but
traffic
of
instead
more
the
In
will, hereafter, use Hsin's
(ATM/C) system
Cantrol
Traffic- Management and
Air
management.
term
used
commonly
"Air Tratfic
Control". [HSI 76]
3. The
proper
timaster-slave"#
with
for
responsibility
all
the
human
to satisfy himself that
reject a
conditions to be met
Thus, we insist that the human controller
in decision making, and tnat
are made without his
explicit approval.
passive, monitoring role for
may somehow exercise a
and presumably
to override that
generate an alternative
the machine to
play an active, dynamic role
no decisions
the decision
a correct recommendation has been
recommendation by requesting certain
recommendation.
should be
the machine, be able to
most importantly be able
will cause
to
accepting
He
information which supports
obtain auxiliary
which
master
decision making.
presented with recommendations by
generated, and
corresponds
relationship
man-machine
the human where he
veto over machine decision making,
intercede on an
we see the human as the
We
exception basis.
Instead
decision maker and the machine as
-19decision
generates
which
system
support"
a "decision
alternatives for him.
1.2
Overview of Previous Research
The topic of ATM/C automation
last
two
A
decades.
number of
has receivea much attention over the
studies have
continuing (see for example [ATH 71],
attempt to
ones from the
[SAR 71], [SCH 73)).
point of view
and are
We will not
ourselves to the
here but will restrict
review all of them
most important
been conducted
of results as
well as the
impact they have had in shaping the cour.se of future work in the area.
Certainly the two most comprehensive studies in terminal area ATM/C
automation have been the Metering and Spacing (M&S) program sponsored by
the
Federal Aviation
Administration (FAA)
development of the
and the
Landing System (MATCALS)
Marine Air Traffic Control and
currently under
development by the US Marines.
The FAA's M&S program ([TAL 78),
Its current version is the third
the early 60's.
test programs.
The first two,
and CAAS (Computer
problems
[TAL 80]) has its origins back in
of a series of field
FASA (Final Approach
Spacing for AXTS)
Aided Approach Spacing), suffered from
(including
procedural
incompatibility,
a variety of
increased controller
workload and in the case of CAAS the lack of a tracker) and thus yielded
(1) The term ATM/C automation will be from here on used to mean
automation of ATM/C decision support as contrasted with data
processing automation which was discussed earlier.
-20simulations using the current version
Computer
few conclusive results.
of the M&S system are being currently conducted at the NAFEC facility in
Atlantic City.
landing capacity
by providing more consistent
landing aircraft
than is now
inter-arrival spacing of
an increase in
attainable, thus assuring
M&S
current
The
utilization.
runway
increase airport
is to
M&S system
of the
primary objective
The
provide for
not
does
design
multiple airports or for multiple runway operations at the same airport.
Provision for departures is made only
stream as
landing
The landing sequence is determined
occurs only
or at
runway
Forward
case
in cases where
one of the
slippage and
of
runway and resequencing of
nominal time of arrival at the
based on the
aircraft
early
the assigned time.
is also
subsequent resequencing
too
arrive at the
an aircraft cannot
intermediate waypoints at
that are
lengthen tne
requests to
manually entered
well as
landing interval by the controller.
landings
through use of normal gaps in the
their intermediate
of
at one
possible in the
waypoints.
The
current
flight plan
the
generator has been
direction,
right
namely
optimal control theory (e.g.
and
faster
path control
navigation capabilities.
has been
the first
marks
M&S effort
accomplished.
away from
time that
an automated
a step in
This has been
complex
algoritnms
based on
[SAR 71] and ESCH 73]) and towards simpler
techniques
consistent with
The objectives and
limited due, apparently,
today's aircraft
M&S program
scope of the
to a decision
by the FAA
that some
-21-
automation system
is a
there
ATM/C
area
Though this is certainly true,
was needed quickly.
term planning of
need for longer
system
automation
which this
a comprehensive terminal
work
attempts
to
lay tne
foundations for.
a quickly implementable system. the
By focusing on
has
are
departures
drawback
flight plans
The
implicitly
only
generated for
standards.
provide altitude separation
aircraft are in conflict.
A closely
of capability to recover and
into
already been
only, and the fact
account).
Another
of conflict resolution.
arriving aircraft are not
future violations of ATC separation
required to
taken
is the lack
current M&S work
in the
which have
of single runway configurations
mentioned (handling
that
two of
drawbacks
some serious
incurred
FAA M&S system
checked for
The human controller is
whenever flight plans
for two
related issue is the system's lack
continue performing the required functions
after controller intervention.
Finally, strict
in
scheduling
efficiency.
adherence to the
runway operations
first-come-first-served sequence
will unduly
reduce the
M&S system's
Optimizing the runway schedule based on the aircraft mix on
hand, was considered but was not incorporated in the current M&S system
since
(1) The FAA has recently sponsored a study to determine the capacity
improvements that can be expected from optimal runway scheduling and
revise the decision if satisfactory levels of improvement are found
possible.
-22...the actual [capacity] improvement [from optimum
[and] can
scheduling] is expected to be less than 3% ...
only be realized when system load is very high. But to
achieve the improvement the sequence will appear abnormal
The abnormal
current ATC practices).
(compared to
sequence will tend to increase controller workload (since
the system's intent will be obscure) and, under heavy
loads any increases in workload or any enigmatic system
performance must be judged inappropriate.''
[TAL 78]
The 3% quoted in this paper is too low an estimate of the expected
improvement
improvements
in capacity.
reductions when
flight
percent
to 15
[DEA 76], indicates that
range
are
in
can result
improvement
achievable.
achieved only
improvement can be
algorithm
frequent changes in
which
is flexible
the schedule of aircraft in
This
dramatic delay
the airport is operating near saturation.
that this
planning
10
capacity
small
relatively
however
the
in
of Dear,
The work
It is true
by implementing a
enough
to accommodate
their initial approach
phase.
It is also
not clear
regard to the obscurity of the
performance.
above statement with
in the
what is meant
system's intent and the enigmatic system
There is nothing enigmatic
about changing the sequence of
operations in order to achieve better runway efficiency.
In fact, final
approach controllers today recognize the efficiency gained by sequencing
similar landing
aircraft
of
whenever
it can
speeds
be done easily.
in direct
succession, and
The optimization
causes the same types of groupings to occur.
do it
of runway schedule
-23-
The
is
(MATCALS)
Landing System
Control And
Air Traffic
Marine
being implemented in response to operational requirements to upgrade and
the
automate
It is intended
The
concept
MATCALS
through
capabilities
all-weather landing
and
control
to be a deployable system,
equipment.
improved
traffic
Control Squadrons (MATCS).
Marine Air Traffic
capabilities of
control
MATCS
air
terminal
designed to replace existing
provide
significantly
advanced
sensors, data
will
automation and
links, displays and operator consoles.
MATCALS
throughout the
and
surveillance
automated
provides
traffic
nautical miles of
airspace within 60
control
the airfield.
In
addition it provides automatic and semi-automatic
landing guidance and
weather conditions down to zero
ceiling for suitably
control under all
equipped aircraft.
MATCALS
capabilities.
its
will
be developed
second
stage
detection
will
The
automatic
third
traffic
and final
with increasing
the early 80's and
ARTS III system.
to today's
be comparable
include
algorithms.
stages each
stage will be completed in
The first
capabilities will
in three
monitoring
stage
will
The
and hazard
be
a fully
automated ATM/C system including such functions as runway scheduling and
flight
plan
generation.
generation algorithms
The
runway
scheduling
flight
plan
will be the prototype
described in this document
software for the third stage of the MATCAL
and
system.
-24-
The
should provide
MATCALS effort
of
problem
concerned with
is
though it
will be of
and experience
MATCALS solutions
operations, the
military
even
automation and,
ATM/C
insight into the
valuable new
great value in the development of an advanced automated ATM/C system for
civil aviation.
AERA
planning
en route
for the
defined in EGOL 81].
Automatic
as
as
well
technology such
Beacon
System
(ETABS).
It
Advisory
and
the
will
matched to
as
true
Service
and
display
the Discrete Adaress
Electronic Tabular
Display Subsystem
four dimensional conflict-free
airspeed.
planning region.
optimal
be used to insure that
the aircraft
Resolution
communications
all IFR flights within the
such
and
beacon provided by
will generate and maintain
profiles, etc.
AERA is
the FAA such as conflict alert,. em
state-of-the-art
(DABS),
characteristics
closely
Traffic
as the mode-S
flight plans for
concept of
AERA will incorporate all the automation functions
metering.
(ATARS),
centralize the
will
The
airspace.
that have thus far been developed by
route
This system. called
Route airspace.
Traffic Control),
Air
En-Route
(Automated
flight
for the En
ATC system
automated
is currently developing an
the M&S system. the FAA
In parallel to
climb
and
Aircraft
descent
projected flight plans are
capabilities.
In
addition AERA will
provide routine aircraft separation and traffic flow control, as well as
clearance generation, delivery and acknowledgement functions.
The AERA concept will
controllers
from
many
improve controller productivity by relieving
of the
routine
functions for
which
they are
-25-
currently responsible.
airspace
enroute
efficient use of
addition it will make more
by limiting
by allowing
and
aircraft
In
traffic capable
movement of
freer
separation of
procedural
need for
the
of area
Finally AERA will coordinate the transition of traffic from
navigation.
the enroute to the terminal area airspace and automatically perform flow
control whenever necessary.
AERA
The
concept is an
planning horizon.
ambitious undertaking
is our hope that the FAA
It
will initiate a similar
year 2000 and
ATC system for the
develop the terminal area
program to
long term
with a
beyond.
1.3
Document Summary
The work described here is part
the
Flight Transportation
MIT
prototype
software
for an
of a continuing research effort at
towards the
Laboratory
automated terminal
designing such a large scale software system, it
to
clearly define
its operation in
distinct but interacting elements.
to understand the ATM/C system
is an
essential
functionally
the
step
interactions
system.
In
is critically important
number of functionally
This breakdown is necessary in order
at the conceptual level.
organization
(or modules).
between various
functional specifications
area ATM/C
terms of a
towards better
related entities
development of
"Finally.
software into
by bringing out
insures
that the
will be compatible
with the
system modules
for each module
of
Furthermore it
it
-26-
of the
operation
The functional
overall system.
breakdown developed
during the course of this research is presented in chapter 2.
of compatibility among the
The issue
apparent in the interactions between the
various system modules is most
man-machine
interface
and
realized
at the
before a
successful design of the
we soon realized
factors affect
considerable research
though we
needed to
be done
man-machine interface were achieved,
that the algorithmic development had
to depend on the
This
means that human
with the
interactions
Even
automation software.
the
outset that
functional specitications of
air traffic
not only each
controller.
module's function but
also the algorithm
This realization led to the development
used to perform that function.
of a real-time interactive ATM/C
simulation facility called TASIM.
The
development effort for TASIM is presented in chapter 3.
TASIM is currently being modified, under
simulate both
enroute airspace.
stations, it will include
controller
software
terminal area and
will
Man-Vehicle
provide
Systems
development by
the
control
Facility
NASA/Ames, [PAR 82].
In
addition to the
pseudo-pilot stations.
air traffic
Research
a NASA/Ames contract, to
In
(MVSRF)
environment
currently
addition to the
the MVSRF includes two cockpit simulators
The new
for the
under
ATC subsystem
and will be used as a testbed
for research in cockpit as well as ATM/C automation.
In parallel
with the development
algorithm for automatic runway
of TASIM, a
real-time heuristic
scheduling was designed and implemented.
-27-
The
can
algorithm
only
it has
however,
Chapter
configurations.
alternative
various
solution
implemented
discussed;
is described
along with
on it by the nature
imposed
This chapter also
display which
runway
the
are
procedures
problem are
formulate the
algorithm
system in which
presents the concept for a
is designed to
which
has
order
been
that are
it will operate.
final approach controller
with the automated
be used in connection
function in
scheduling
presented and
particular requirements
of the
runway
single
of runway scheduling;
4 discusses the problem
ways to
for
tested
successfully
been
To date,
configurations.
multiple runway
handle
delivery of
precision
to allow
landing aircraft at the runway in the absence of flight plan automation.
The final part
flight
plan
of the research addressed the
generation.
The
general
problem of automatic
approach has
been
that first,
flight plans must be compatible with conventional navigation capability,
and
second,
preserved
flight
to
plan
flexibility
the greatest
in
order
Accordingly, flight
generated
based
on
structure should be
at each
in
to
Based
provide
possible at
maximum
prespecified
shortening
every point
a number of linear
ground
One possible
on that we develop the
must be
along the
rescheduling flexibility.
track
capable of providing a number
intermediate point.
and analyzed.
extent
plans consist of
a
and
path stretching
legs, and are
structure.
ThIs
of alternative paths
structure has been developed
methodology for the design
of a traffic path generation algorithm which is presented in chapter 5.
-28-
Chapter 6
summarizes the results
and conclusions ot
identifies topics for further research and development.
the work and
-
blank page -
-29CHAPTER 2
FUNCTIONAL DESCRIPTION OF AN AUTOMATED TERMINAL AREA
ATM/C SYSTEM
2.1
Introduction
system components
The
aggregation.
described
be
can
system
complex
Any
at each
various
at
level can
of
subsystem
Control
system
or
of components
number
a
into
primary
system.
will be
Finally
subsystems.
in this
interest
each of
broken down into
themselves be
be described and analyzed from this
We will begin with the overall
point of view.
of
In this chapter
complex systems requiring the same type of description.
the terminal area ATM/C system will
levels
system and subdivide it
We next
research,
the elements
focus
namely
of the
on the
the Ground
Ground Control
and its operation
functional modules
will be discussed.
At each level we will focus not only on the function of each of the
components
but also
on the
that is required, as well as
to the others as a result
control
system
response
therefore develops the
system and
interactions, the
the flow of information from one component
of specific external events (e.g.
to
a
conformance
alert).
This
the ground
chapter,
framework for the design of
the automated ATM/C
at a top level,
of the requirements
for the understanding,
and the purpose
exchange of information
of each of the automation functions.
At the same time
-30it
sets the
basis for
since the
software,
of the automation
and design
the organization
functional modules in
breakdown can be
the final
regarded as top level software modules for the ATM/C system.
2.2
The Generic Terminal Area ATM/C System
Air
interactive
system.
Management
Traffic
system
that
Control
and
can be
modeled as
best
is a
(ATM/C)
complex,
feedback control
a
Its elements can be grouped into six generic subsystems:
Aircraft Control system
(A/C)
Air-to-Air Data link system
(AA COM)
Automatic Ground Control system
(AGCS)
Ground-to-Air Data link system
(GA COM)
Air-to-Ground Data Acquisition system
(AG COM)
(GG COM)
Ground-to-Ground Data link system
An overall block diagram of the terminal area ATM/C system is shown
in
figure 2-1.
The major
(AGCS) and the Aircraft Control (A/C)
feedback
control
control
loop for
systems themselves,
ATM/C.
the Automatic
elements are
The
Ground Control
While each of these are
systems.
they are
Aircraft Control
elements of
the major
systems (or "targets")
output their "state" vectors, P.(t) and P.(t), which are measured by the
.1
Data Acquisition (AG COM)
J
4
system to serve as primary
dynamic input data
(1) The material in section 2.2 is drawn from the work of Hsin [HSI 76]
where the reader is referred for an excellent detailed discussion of
advanced ATM/C systems.
-31-
Ground Control
Sectors
GA
GA
COM
F
4
F.
I
AGCS
js
J
IAG
COM
4.
-
-
GG COM
OTHER
ATC
FACILITIES
Figure 2-1.
Overall Schematic Diagram for the Terminal Area
ATM/C System.
- - i
-32-
to
the
i.
aircraft
processes this
AGCS
The
on the
information
P.(t) includes
AGCS.
past
positions of
and
produces its
intormation
output, F. and F., which are ATC commands or intended flight plans which
1J
describe the future path of the aircraft.
dynamic inputs to
These in turn are the primary
the Aircraft Control systems.
various Data link
The
systems accomplish the required flow of information.
and
availability
The
importance
distribution
ATM/C system.
operation ot the
in the
information
of
is of
key
It consists or the
current "tstate", P.(t), and the "intended" future state or flight plans,
F., of the aircraft targets in
P.(t) includes the aircraft
the system.
1
1.
position
and
altitude
as
well as
heading,
bank
angle,
etc.
The
available.
to
Aircraft
the aircraft
aircraft
ATM/C
velocity, accelerations,
is not globally
information
heading, for example, is currently available only
control system
itself.
Due
to the
variation in its
completeness and because of measurement errors the aircraft state vector
takes three distinct forms:
1. The actual aircraft state, P.(t)
2. The ground
information
measured aircraft state,
on
target i
available
Ground Control system, and
P is(), is estimated
to sector
s
of the
includes the errors introduced
by the Air-to-Ground Data Acquisition system.
.t),
3. The airborne measured state, P.
11
information available to the
is the aircraft state
Aircraft Control system, and
-33-
includes
be available to
in the system can also
of other aircraft
aircraft
Information on the state
A/C.
guidance subsystems of the
navigation and
the
by
introduced
errors
the
i, either
P.. Ct)
Ji.
form.
absolute
in
or
in
Absolute state vector inrormation
relative form, AP. .(t).
may be relayed to other aircraft through the ground to air
be obtained by
relative information may
link while
data
separation assurance systems onboard aircraft i.
The
discrepancy
between
measured
aircraft state
requires
some
further
difference, (i.e.
their
accuracy,
accuracy
their topological
timeliness,
is obvious
different systems.
the airborne
period
airborne
and
they
since
Also by virtue
In
state
quantities. (e.g.
to
of
information
their defining
availability), they also differ in
completeness.
are
difference in
Their
produced
measurements
by
two
of the measurement systems involved,
instantaneous while ground
at relatively low sampling rates
of revolution
the airborne
reaching consequences and
addition
state information is generally
measurements are made
the
information has far
elaboration.
and
measured
ground
the
antenna.
the radar
includes
a
wide
position, altitude, airspeed,
which depend on
Finally,
variety
of
while the
measured
vertical speed. etc.)
(1) There exist, of course, military "tracking" radars which can
directly control the direction of the beam thus producing very high
These radars measure the
sampling rates for particular targets.
than conventional
accuracy
better
aircraft position with much
will be used for
these
that
radars. It is very unlikely. however,
future.
civil air traffic control in the foreseeable
-34today's
surveillance
receive
encoded
transponders.
altitude
and
identity
and can
from
mode-C
information
on the
information
the airborne
therefore,
Generally
and bearing
only range
systems measure
aircraft state is more complete and in many cases more accurate than the
intormation available to the ground.
The
issues
and
Sophisticated
and
accuracy
the
feasibility
the
in
connection
achieved
plan
flight
automated
with
be
system.
without high
These issues will be discussed
quality surveillance and tracking data..
again
ATM/C
automated
cannot
ATM/C
in
automation
determine the
are factors which
of
state
aircraft
of
availability
important since these
information are
efficiency
of
generation
and
conformance monitoring.
We can now show the elements within the block for the AGCS shown in
figure
2-1.
functional
As
depicted in
elements:
the
figure
Flight
the AGCS
2-2,
Plan Generator
has
(FPG),
three major
the Traffic
Monitor (ATM). and the Command Processor (ACP).
The FPG is responsible for creating decisions regarding the flow of
traffic
in the
terminal area
ATM/C system.
The state of all controllea aircraft
terms of an optimization process.
is the
input variable
of the
process.
commands which define a flight plan
i.
The
ATM/C
rules
and
It is best described in
,
Thegoutput
will be
the ATC
F., for each controlled aircraft,
regulations,
and
the
aircraft's
dynamic
(1) The term "flight plan" is formally defined later in this section.
-35-
ATM/C
TRAFFIC
MONITOR
(ATM)
ATM/C
COMMAND
Ps
. S
PROCESSOR
(ACP)
ATM/C
FLIGHT PLAN
GENERATOR
(FPG)
-Conflict Checking
ATM/C DISPLAYS
(TD and CTID)
Figure 2-2.
The Major Elements of the AGCS.
-36-
minimum
(e.g.
characteristics
the
determine
constraints
which
decisions.
Finally,
maximum
and
set
of
etc.)
airspeeds,
feasible
ground
are
control
the objective function
the ATM/C strategy defines
which the FPG is to optimize.
two sub-elements:
Monitor naturally divides into
The Traffic
the
Conformance monitor and the Hazard Monitor.
The
(CD)
Conformance Monitor,
consisting of
the Conformance Detection
and the Conformance Resolution Path Generation (CRPG)
in
preventive
nature; it
control decision and generates a
the
established flight
The
Hazard Monitor,
of
plans exceed
satety
Detection (RD)
rules; it
between the Hazard
Monitor and the
and the
is responsible for
generates a
whenever violation of ATM/C standards is imminent.
be drawn
the ground
to
prespecified conformance limits.I
the Hazard
consisting of
the ATM/C
is-
conformance alert when deviations from
Path Generation (HRPG) functions,
Hazard Resolution
enforcement
aircraft adherence
monitors
functions,
hazard alert
A distinction should
conflict check.
"Hazard"
implies a perilous current situation in terms of the actual positions of
aircraft, whereas "tconflict" refers to a possible future situation based
on
flight
plans.
In
particular, a
flight plan
is in
conflict with
another if an aircraft adhering to it within the conformance limits wiil
create a hazard (i.e.
violate the
ATM/C rules) at some future point in
(1) The conformance limits depend primarily on the FPG logic, but the
accuracy of the surveillance and the onboard navigation equipment is
also a factor.
-37Accordingly the conflict check is an integral part of the FPG and
time.
appears
2-2 to
figure
in
generated are
plans
that flight
indicate
conflict-free.
Processor also subdivides into
The Command
the Command
(CG) and
Generator
Command
plan.
for
aircraft will
followed, the
same flight plan may
The
aircraft
different
if the commands
its flight
conformance to
remain in
conceivably produce different commands
the
aircraft
navigation
pilot can be
of commands the
what type
determine
since
The Command
Activator (CA).
times of initiation, such that
commands and associated
the
determine for each aircraft a set of
Generator uses flight plan data to
are
two sub-elements:
capabilities
expected to accept.
The Command Activator is responsible for the timely dispatch of commands
to the air traffic controller and (if digital data link is available) to
the aircraft.
The
human
only manual
controller who
controls the
real time inputs to the Traffic
Information Display
form
of computer
decision
support
controller.
actions.
He
in figure
2-2 is that of the
overall system
performance through
activity depicted
(CTID)
Display (TD) and the Controller Tabular
software.
algorithms which can
within
remains
the
the
All
other elements can
have the
operate automatically providing
framework
established
commander responsible
by
for
the
ATM/C
the system's
-38Before we discuss the various
useful in describing flight plans.
the following definitions will be
point
3-dimensional space
W in
W(t),
define tim-point.
automation functions in more detail,
callea a waypoint.
will be
dimensional waypoints for
as four
A
We shall
which the
is
flight plan for aircraft i. F.,
3.
time dimension is also specified.
A
a set ot waypoints or time-points
detining a path in 3 or 4-dimensional
limit our
We
include
only time-points
(t tf).
of a
definition
space.
coordinate lies
whose time
plan to
area flight
terminal
in some interval
interval of interest will be the
Unless stated otherwise, the
time during which the aircraft is under the control of the terminal area
ATM/C controller.
it
is handed off
of
runway, i.e.
the
exit
aircraft's
to the
from
interest is
the
).
departing aircraft the
For
the scheduled
of arrival
time
F. is feasible
if it is consistent
characteristics
(such as
i
2.
performance
at the
time of arrival at
takeoff time, until the scheduled
fix (SXT.).
time of
the scheduled
controller to
ATM/C
assigned runway n (STAR
arrival at the
interval
For a landing aircraft i, this would be from the time
airspeed,
with the
climb and
descent rates, taxi speeds, etc.).
2.3
The ATM/C Flight Plan Generator
In this section the FPG is described in more detail with particular
interactions between the
emphasis
on the
that are
identified.
In figure 2-3
three distinct functions:
various automation functions
the FPG is shown
as consisting ot
-39-
j
i Pjs
r
D
COMMAND
ACTIVATOR
(CD)
2)j
COMMAND
F.
GENERATOR
DETECTION
ION
7
(CA)
CONFORMANCE
I
R
D
ION
CONFORMANCE
RESOLUTION
G
ION
PATH
GENERATION
(CRPG)
(CG)
L
LCommands
f.azard Avoidanc
Flight Plan
-
i[Re-acquisition Commands
--
-
-
-
-
-
-
FI Res cheduling
I
TRAFFIC
FLIGHT PLAN
NOMINAL
FLIGHT PLAN
RUNWAY
GENERATOR
SCHEDULER
(TFPG)
(RS)
4-
I
1P.
is
GENERATOR
(NFPG)
IJ
-Conflict
Checking
ATM/C DISPLAYS
(TD and CTID)
Figure 2-3.
Functional Block Diagram of the AGCS.
P.s
Pis
.%7
'js
-40-
1. Nominal Flight Plan Generator (NFPG)
2. Runway Scheduler (RS)
3. Traffic Flight Plan Generator (TFPG)
Nominal Flight Plan Generator
2.3.1
Given the
present position, destination,
characteristics, the NFPG determines an
would be
(terminal
plan assignment assumes
area speed,
could
for this purpose.
be used
earliest times of
the airport.
arrival at
,
that
any other traffic.
nominal performance characteristics
descent or climb
Where applicable,
aircraft type.
efficient flight plan. F
in the absence of
assigned to an aircraft
The flight
and aircraft performance
rates etc.)
depending on the
standard deparrure and arrival routes
The nominal
flight plan establisnes
arrival at various points in the
terminal area or at
Of particular interest is the earliest (preferred) time of
the runway threshold
(first-come-first-served)
(PTAR)1 which establishes
sequence of
operations (NSAR).
the nominal
NSAR is the
basis for the sequencing constraints which will be introduced in section
2.3.2
The NFPG is invoked whenever
for
takeoff or
for landing)
schedule change (e.g.
new aircraft enter the system (either
or when
the ATM/C
controller requests a
in the case of a missed approach).
(1) The term "time of arrival at the runway" is used for both landings
For landings it is the time the aircraft crosses
and departures.
the runway threshold, while for departures it corresponds to the
time the takeoff roll starts.
-41-
2.3.2
Runway Scheduler
Given a runway system consisting of Nr runways, and Na aircraft, we
the set of STAR.in's
runway schedule S as
define a
(scheduled times of
arrival at the runways) for all aircraft i and runways n in the system.
S = { STAR.in
-
Adoption
a
of
,
specific
scheduling
decision.
assignment
of
}
i=1,2,3,...,N ar n=1,2,3,...,N
schedule will
runway
Note
runways to
that
a scheduling
aircraft
be
referred
decision
whenever more
all m other than n. By convention we
set STAR.
not the
assigned runway for aircraft
i. When two or
active,
we may
restrict landing
as a
includes the
that one
active. If runway n is assigned to aircraft i, STAR.
to
runway is
has no meaning for
to zero if runway m is
more runways are
operations to certain
and/or takeoff
runways, perhaps depending on the type of aircraft.
The
ATM/C
runway schedule
system (i.e.
defines the
efficiency of
the time interval
the terminal area
between successive operations).
Generating the schedule thus represents the major optimization effort in
the
AGCS decision
process.
objectives and strategies
Unlike
optimized,
system
are
and strategies
well defined
performance criterion
system wnere ATM/C
employed to achieve them are
implicit and qualitative terms,
be
today's manual
the objectives,,
which are
functions
to be optimized
only defined in
performance criteria to
implemented in
of the
STAR's.
For
might be average
the automatic
example, the
aircraft delay
-42and
the
strategy
accomplishes
performance
adopted will
that.
In
criteria,
be
chapter
their
a specific
4 we
discuss
mathematical
runway
in
schedule which
detail alternative
representation
and
their
effects on runway scheduling algorithms.
The Runway
Scheduler is a
defined optimization problem:
computer algorithm which
solves a well
Given the nominal sequence at the runways
(NSAR), and for each aircraft in the system:
1. its performance characteristics,
2.
its
Earliest Feasible
Time of Arrival,
EFTAR. , at each
in
active runway n, and
3. its
Latest
Feasible Time
ArrivaL, LFTAR. ,
in
of
at each
.
1
active runway n
find a particular schedule S
which optimizes the performance criterion
and satisfies the following constraints:
1. Spacing:
the minimum
required time
all pairs of operations i and
(1) EFTAR.
in
and
LFTAR. will
in
flight planning logic.
and 5.
depend
j are
separations between
not violated;
on the
approach routes
and the
They will be discussed further in chapters 4
-432. Sequencing:
no
aircraft
is
shifted
prespecified number of positions
by
more
than
a
from its position in the
nominal sequence;
3. Scheduling:
for
the
scheduled time of arrival
all landing
time interval,
aircraft should
at the runway
be within
i.e., for all landing
the feasible
aircraft i assigned
to runway n,
EFTAR.
in
Ldie:
4.
The
<STAR. < LFTAR.
in
in
current scheduled time of
arrival at the
runway cannot change if it is within the prespecified lead
time from the current clock time;
Lead time constraints are designed
the schedule.
For departing aircraft
smooth operation of the takeoff
constraints
notice
also
of their
stabilizes as
absolutely
they allow for
taxiing time and
queue ne-ar the runway threshold.
provide pilots
of
exact landing time
landing can be made.
to avoid last minute changes in
landing aircraft
so that
These
enough advanced
adequate preparations for
Under normal conditions the STAR for each aircraft
its scheduled time approaches.
ensures that
time interval before STAR.
this will
The lead time constraint
always happen
at some prespecified
-44-
2.3.3
Traffic Flight Plan Generator
Given the
performance
new runway schedule
*
S
aircraft in
all the
characteristics of
positions, and the
the current
,
the terminal area
the ground control decision process by
ATM/C system, the TFPG completes
generating a new Flight Plan Decision,
*
*
{ F.
((t)
,
i=1,
2
,
3
,...,N
}
i.e. 4-dimensional flight plans F., such that:
for aircraft i, i=1.2,...,N a
1. F. is feasible
3.
2. minimum
airborne
satisfied
separations are
throughout, i.e.
the flight plans are conflict-free.
*
3. a smooth transition between F. and F. is provided for
all i=1,2,...,N
a
4. The pilot workload from the time of system entry to the time of
system exit is kept within reasonable limits.
The
since
flight plans
generated here
conflicting traffic
above constraints.
are called
is taken into
traffic flight plans
account by the
second of the
-45The TFPG
after a new
is normally invoked
scheduling decision has
a new aircraft entering the system or
been made due to
due to a missed
It may also be invoked however due to a conformance alert.
approach.
possibility that a
is a
There
free flight plans
set of conflict
In this case the TFPG will invoke the runway scheduler
cannot be found.
requesting a schedule modification.
2.4
ATM/C Command Processor
This function consists of
two sub-functions, the Command Generator
(CG) and the Command Activator (CA).
guide the aircraft along its assigned
commands which, if followed, will
flight
The
plan.
specific
commands take the
commonly
terminology
flight plan
used
today
for
VOR/DME
or
TACAN
Navigation).
commands is of
and altitude.
capability
navigation)
or
the prevailing winds.
is either
advanced
For conventionally equipped
be
a
the
The
conventional (i.e.,
(i.e.,
3D
or
4D
Area
aircraft the set of allowable
the "radar vectoring" type whicg
Aircraft
Ci)
on
generated depend
(ii)
For
communications.
voice
the actual commands
navigational
alphanumeric messages in
form of
aircraft's navigational capability, and
aircraft's
a set of
plan data to generate
Generator uses flight
The Command
equipped with 3D or 4D
specify heading,
speed
Area Navigation (RNAV),
can be commanded to track directly to the next waypoint or time-point.
-46-
In order to achieve good conformance to 4-dimensional flight plans,
speeds
and headings
in the
winds
vicinity of
the aircraft.
arrival routes for
In that
wind is performed
particularly in the early
late stages of
case a command to
by the pilot (or autopilot)
will be preferable
segment of a
Occasionally a
track a VOR
since compensation for
preferable to a heading command
radial would be
for the estimated
landing aircraft, and the
for takeoffs.
departure routes
RNAV commands
to account
coincide with a VOR radial,
flight plan will
stages of
should be corrected
in tracking.
to heading commands and
Similarly,
will be used
whenever possible.
Each command
message includes at
least four additional
pieces of
information necessary to its further processing:
1. the
identity
of the
aircraft
to which
the
message is
directed
2. the time of issuance to the ATM/C controller
3.
the time of transmission to the aircraft
4. the acknowledgement of the command by the aircraft
The difference between issuance and
as to give the ATM/C controller
transmission time will be such
the opportunity to study and (if deemed
necessary) modify the command, to validate its automatic transmission to
-47the aircraft at the correct time if digital data link is used for ground
to air communications, and to prepare for voice transmission otherwise.
The Command Activator is responsible
command messages.
future commands for
This list of
First, the controller will be able to
When the
will be moved
for
to
under his control.
CTID and is updated when
The controller may also modify any command at this
command issuance time
has been
to the issuance area which may
Additional methods to
easier reference.
attention
any or all aircraft
commands can be displayed in the
flight plans change.
stage.
commands to the ATM/C controller
The presentation of
can be achieved in three stages.
review all
for the timely dispatching of
it (e.g.
blinking) may
be
reached, the command
be in the traffic display
attract the controller's
used.
Finally
after the
transmission to the aircraft has been initiated, the command is moved to
a
post
Thus the
view area.
controller is reminded of
commands for all aircraft under his control.
the "active"
In this area, the commands
may also be flagged when acknowledged by the aircraft crew.
The transmission
validates the
capable
of
to the aircraft is initiated when the controller
command.
both
voice
Since it can
and
data
be assumed that the
link
(mode-S)
AGCS will be
communications, the
processing of the validated command by the Command Activator will depend
on the aircraft communication capabilities.
The important issue
in this respect is that
not be responsible for determining
the controller should
the transmission method.
This means
-48-
mode-S
capability.
area.
If not,
voice
workload
of
associated
with routine
commands to
this
to alleviate
regard
is the
the controller
of commands
transmission
may be able
The computer
transmit voice
option in
synthesizers
area and the
command through
transmit the
has to
interesting
using voice
communications.
and
An
communication.
possibility
that he
to the postview
the issuance
will remain in
the command
target aircraft has
the
will be moved
so. the command
If
will realize
controller
if
Activator will determine
the Command
that
via voice
to automatically synthesize
all aircraft
that do
not have mode-S
capability.1
If validation is not made by
the
aircraft,
message to the
the Command
the specified time of transmission to
Activator
ATM/C controller and no action is taken until the ATM/C
controller validates the clearance or
aircraft to follow.
transmitted to the
without
messages
appropriate alert
dispatches an
initiates another command for the
Similar processing takes place after the command is
aircraft.
acknowledgement
of
are automatically
If a specified time
the
command
dispatched to
by
interval has elapsed
the
pilot,
the pilot
and to
appropriate
the ATM/C
controller.
(1) Voice synthesis has tremendous potential in many
automation. The technology, however, is still in
lot of research is still required with regard to
as well as to the human factors aspect before its
assessed.
aspects of ATM/C
its infancy and a
the technological
usefulness can be
-492.5
The ATM/C Traffic Monitor
Conformance Detection and Resolution
2.5.1
primary mode
is the
This
monitoring for
of automatic
of the aircraft
the indicated surveillance position
Deviations between
the AGCS.
and the desired position according to its currently assigned flight plan
are
When deviations
monitored.
(which may
respect
to
on the
depend
assigned
the
established limits
exceed externally
runway)
the
location of
geographic
the
out of
is declared
aircraft
alert is generated.
flight plan and a conformance
conformance with its
aircraft with
This will generally cause the TFPG to be invoked.
In some cases
back
it will be reasonable to
into conformance
is performed
function
Generator
Path
Resolution
alert will
flight plan
are detected.
and the
the
This
of
type
Longitudinal deviations,
recovery
from
a
deviations from the
on the other hand,
between the aircraft's
for the
to the fact that no aircraft
its nominal airspeed,
This
the Conformance
by
TFPG
when lateral
nominal airspeed assumed
In addition
fly at exactly
of
be the result of discrepancies
will most likely
planning.
(CRPG).
be desirable
conformance
airspeed
"correction" vectoring.
through immediate
independent
quickly bring the aircraft
purposes of flight
can be expected to
such disagepancies
will occur due
to erroneous estimate of the wind speed and direction particularly after
a change in
heading.
In that case, it will
generally be preferable to
-50aircraft based on
plan for the
the flight
modify
the new information
(namely the new estimate of the ground speed) available to the TFPG.
2.5.2
Hazard Detection
generation of conflict-free flight
separation assurance, namely through
out of
aircraft must be
Controlled
conformance monitoring.
plans and
of providing
primary mode
to the
backup
is a
Detection
Hazard
conformance before they can be in hazard.
the snort term straight line
Traditionally, this function monitors
from
separation
positions
in
other aircraft
of
"unassuming" mode (U-mode).
in
particularly
the
the system.
We
aircraft
projections
will call
of the
this the
This approach has not been very successful,
area where
terminal
the
as
as well
terrain
ground
of
seconds)
30
of
order
the
of
(typically
projections
the
proximity
of aircraft
results in high false alarm rate.
provided to the
We
will call
rate.
There
is a certain
either by
this method
danger to
is primarily
responsible for
situation
to detect such
will be
biased by
information
may be
to reduce the false
mode (I-mode).
the "informed"
the I-mode
the air traffic controller
particularly suited
the
plan
Hazard Detection function in order
alarm
function
system, flight
ATM/C
automated
an
In
since the
Hazard Detection
safeguarding against "blunders"
The
I-mode is not
blunders because its
assessment of
or the pilot.
the -fact that
it
"knows"
what the
-51It is possible, however, to maintain the same
aircraft should be doing.
same time to
blunders and at the
performance with respect to
level of
increase the reliability of the Hazard Detection function by having both
modes operating in parallel.
can
We
when the
declared
it will
that
have
be resolved
flight plan
according to
level is
the I-mode determines
hazard but
U-mode declares a
first
The
alert.
hazard
of
two levels
inrormation.
For
example, one of the two aircraft involved is due to initiate a 90 degree
In this case the controller may be warned
turn in the next few seconds.
expected command is actually in the process
in order to insure that the
of
No
initiated.
being
by
requested
declared when both the U-mode and
unless explicitly
second level
will be
the I-mode declare a hazard,
or when
The
controller.
traffic
the air
is taken
further action
the (internally determined) probability of an actual blunder has reaced
question
is to
addition
to
conformance
the
hazard
be initiated
hazard
monitoring
Hazard Resolution
resolution
that the command in
value (for example, the time
a specified threshold
has
message
for the
been
aircraft
Path Generator will
commands.
the
to
Note
In
reached).
traffic
air
will be
be invoked in
that
the
this
case, in
controller,
suspended,
and the
order to generate
conformance
detection
function should be declaring a conformance alert and trying to solve the
problem at the same time.
the conflict checking
(1) This will generally not be possible if
function of the Traffic Flight Plan Generator is operating properly.
-52-
At this time
any more.
to be in hazard
found not
until it is
alert status is maintained for the aircraft
A hazard
the hazard alert is
If there had been a second level alert, conformance monitoring
removed.
Presumably
is resumed.
out of
aircraft
have placed the
the hazard avoidance path will
plan.
original flight
conformance with its
A normal
conformance alert to will then be generated causing the CRPG or the TFPG
to
be invoked
in order
the problem
to solve
the A1'M/C
of returning
system in its primary monitoring mode.
Hazard Resolution Path Generator
2.5.3
function
This
occurs.
is invoked
modes of operation
Two
a second
whenever
hazard alert
level
the type of
are possible depending on
ground-to-air communication available.
If voice communication
will
be
generated
and
processing
and made
recommended avoidance path
is employed, a
to the
available
immediate transmission
Command
the ATM/C
to
Generator for
controller.
The
controller then chooses to validate the recommended solution or generate
his
own.
In
this mode
it may
be advantageous
to provide
the ATM/C
controller with more than one alternative.
If a
feasible
digital
generate a
to
processing by
This
mode
ground-to-air data
single
is available,
avoidance path
the Command Generator
insures
link
faster reaction
and, request immediate
and transmission to
time
to the
it may be
the aircraft.
imminent
hazard but
-53-
paths.
reliable
a very
requires
algorithm for
generating
hazard avoidance
The hazard resolution commands are also transmitted to the ATM/C
controller to avoid conflicting actions on
his part.
He is not however
required to take active part in the process of resolving the hazard.
He
can participate in returning the aircraft into conformance.
absolutely necessary element
is not an
The HRPG
system and, at least initially,
of the automated
the ATM/C controller could perform this
function manually without any loss in the system performance.
2.6
Traffic and CTID Displays
display called
traffic
uses
ATM/C controller
human
The
the Traffic Display
information called
two displays:
(TD), and an
the Controller
a
normal radar
auxiliary display for
Tabular Information Display
(CTID).
component of the man-machine interface
The displays are the output
of
the
system
AGCS.
state (current
displayed on
pertinent information
display all
can
They
and performance.
and future),
either display is controlled by the
real-time inputs on the associated
about the
The information
ATM/C controller via
keyboard or other data entry device.
Information nominally displayed on the TD includes:
1. movements
of
aircraft
surveillance area.
a
short
targets
For every
alphanumeric
block
within
the
radar
aircraft displayed there is
containing
the
aircraft's
-54-
its altitude,
identity,
current position
its
aircraft)
landing
estimated ground
speed and (for
the landing
in
sequence;
such
data
2. geographical
as
etc.,
obstructions.
as
location
the
well
the
as
of
airports,
navigational aids
located within the surveillance area;
3.
well
controlled aircraft as
to be transmitted to
commands
as other
such as
priority messages
conformance or
hazard alerts.
The CTID
will display information
need for reference.
that the ATM/C
but does not require constant
controller will
cognizance of.
information includes:
1. All pertinent information on aircraft in the system.
takes
two
(i)
forms:
constantly
displayed
This
"summary"
information including aircraft identification, transponder
code,
current
information
displayed
explicitly requested
description
of
and (ii)
more detailed
particular
aircraft when
clearances. etc.
the
about a
by the ATM/C
currently
controller (e.g.
assigned
flight
full
plan.
requested landing speed or takeoff weight for the aircraft
etc.);
Such
-552.
System resource
data such as condition
of active runways
and taxiways, weather data, etc.;
3.
System performance
data such as the
statistics on runway
utilization, average delays etc.
2.7
The ATM/C Controller's Role
The ATM/C controller
be the "policy
and
the
is the human element in
maker" or "chief executive" who
general
framework within
which
the AGCS.
He should
determines the strategy
the software
is allowed to
generate the plan of action down to the very small details.
He controls the decision process in the following ways:
1. overrides the computer generated runway schedule,
2. imposes
additional constraints
process.
Such
aircraft
land as
on the
constraints include:
soon as
emergency), require that an
runway scheduling
requiring
possible, (as
in the
that an
case ot
aircraft maintain its current
position in the sequence. etc.,
3. requests changes in the configuration (e.g.. change in the
active runways, change in ATM/C strategy. etc.);
4. validates commands for transmission to various aircraft;
-56-
computer
5. Edits
generated
and
commands
initiates their
transmission process.
controller's request is acknowledged and he
In all cases the ATM/C
is informed
on the
the request
results of
information is
when such
necessary to insure that the intended action was taken.
The general
philosophy followed here is that the ATM/C controller
does not simply monitor the computer performance but remains responsible
for safe separation and expeditious flow of traffic, and must be totally
in control of the system.
The automated system must be a tool to assist
the ATM/C controller in carrying out the task safely and expeditiously.
The
to
extent
which
automated
system
will
be
successfully
implemented in the future depends on the success or failure to develop a
suitably designed man-machine interface that allows the ATM/C controller
to be
at all times
in control of
capable of exerting such tight
in a predictable
the
found
same time,
between
during normal
operation a
controllability and
extensive interference
of ATM/C automation.
will be required betore
this problem can be reached.
point of view.
delicate balance
options and extensive testing of
interactive control
and, if desired,
control that the computer programs react
manner from the ATM/C controller's
completely defeat the purpose
of the available
the overall system
At
should be
which might
Careful examination
different methods of
a satisfactory solution to
-57CHAPTER 3
A SIMULATION TESTBED FOR THE DEVELOPMENT OF ATM/C
AUTOMATION SOFTWARE
3.1
Overview
to deal again and again
of this work we were forced
In the course
with "human factors" aspects of the problem of ATM/C automation in spite
of
the
that the
fact
a fast time simulation.
be tested using
in, all
cases algorithmic~ and
account
the method
test
effectively
described
area
early
apparent
in the
and
It was concluded finally that
the ATM/C
demonstrate
previous chapter,
any
the
of
necessary.
It is a substantial
of the
extension
controller.
order to
in
automation
It was
functions
time, interactive terminal
a real
A terminal
was designed and implemented
facility, TASIM,
has to -take into
effort that,
research
in the
facility was
simulation
algorithms which would
software develonment
of interaction with
and deveiop
study
was to
and flight plan generation
automatic scheduling
therefore
intention
original
area simulation
to fulfill this purpose.
original
work done
by Heinz,
[HEI 76].
TASIM captures in great detail the performance of all elements or a
terminal
area
ATM/C system.
Particular
emphasis was
placed
on the
interactions between various elements and on the flow of data within the
ATM/C system.
To insure correct
and easily identifiable program logic,
-58a
top-down
design
approach to
techniques were
independent functional
clearly
software
defined
used to produce
adopted.
a system
parts (software modules)
interfaces and
The final product is a
development was
Composite
consisting of highly
interconnected through
performing clearly
derined functions.
realistic and flexible simulation facility which
can be responsive to the needs
provides a vital research tool
of human factors research.
for the development
It therefore
of ATM/C automated
decision support.
This chapter will
the features
3.2
that are of
within
enironmernnt
give a brief description or
wh1,ti
TASIM and highlight
consequence in realistically
ATM/C
uoain
ucin
reproducing the
h!2ve
to
efom
The Hardware Environment
The current hardware
diagram for TASIM is shown
in figure 3-1.
A
VAX-1l/780 provides the main computing power and houses about 95% of the
software.
A PDP-1l/34 computer is used to generate the traffic (radar)
display
and
receive
information
communicates via
on
active
data
link with
aircraft
in
the VAX-l1/780
the
only to
simulation
or
on
controller inputs requesting some type of display restructuring function
to be performed.
In addition the facility currently
includes an ATM/C
controller station and a simulation monitoring station.
The
ATM/C controller
computer as the TD
station uses a
MEGATEK calligraphic display
and a VT-100 video terminal as the
CTID.
It is the
VAX-11A780
PDP-1/34
Figure 3-1.
Current TASIM Hardware Diagram.
-60primary man-machine
TD in
ARTS III format,
Display)
is
used
generated by
data
interface.
entry
The traffic movements
while the CTID
to
display
(Controller Traffic Information
supplementary
various automation functions.
device
allowing
are displayed on
The
tabular
information
VT-100 keyboard is the
real-time interaction
between
the ATM/C
controller and the simulation software functions.
The
simulation monitoring
station, consisting of
video terminal,
is currently used
well
and
as
start
end
a second VT-100
only to initialize the
the simulation
run.
Since
data base as
it
also allows
real-time interaction with the simulation software, its capabilities can
be expanded
real
in the fvlture to
time.
levels,
This capability
weather and
includp modification of the
data bqse in
will
over traffic
allow direct
wind conditions,
unusual situations that merit
and can
control
be used
to bring about
close investigation (e.g.
procedures for
changing runways due to shifting wind direction, etc.).
In the near
SANDERS
future TASIM will be moved to
Graphics-7
displays as
TD's and
a VAX-ll/750 with three
three Texas
Instruments 940
terminals as CTID's, [PAR 82].
3.3
Major System Components and Interfaces
The
processes,
simulation
as
shown
software consists
in
figure
3-2.
"stand-alone" computer program operating
of
4 major
Each
components, called
process
is
a
separate
in parallel and independent of
Figure 3-2.
TASIM Software Processes.
-62the
A common, memory-resident
the message.
possibly the process which sent
mailbox, a
message contents and
on the
function is performed depending
specific
in the
a message is received
it. When
with
associated
a mailbox
has
process
Each
mailboxea.
called
devices
the others via
process communicates with
VAX computer each
I/0
virtual
the VAX-11/780 CPU.
PDP-ll/34 computer, they all share
operates on the
Within the
display program which
the traffic
the exception of
With
others.
data base is shared by all VAX-1l processes.
the simulation execution.
Mnnitor Proce.ss oversees
The Simulation
Prior to the beginning of real-time operation, it initializes the global
for
responsible
the
required.
processes
as
displays,
statistics
and allocation
timing
on
In
addition
CPU
usage.
the execution
During
execution.
the real-time
to start
input
it
and
it is
time
to various
and
optionally
CPU
of
collects,
the
subordinate
accepts operator
mailboxes, and
the communication
creates
processes,
of
loading
proper
the
insures
and
initializes
data,
status
of
all
other
simulation processes.
The Position
updating
the
Generator Process is responsible
state (e.g.
position,
active aircraft in the simulation.
of the
Aircraft Control Systems
System
(surveillance
overall
2-1).
schematic
radar and
diagram of
It
for maintaining and
speed, altitude.
etc.)
of all
provides realistic dynamic models
and the Air-to-Ground
tracking
the terminal
Data Acquisition
processing) depicted
area ATM/C
in the
system (figure
-63The
Process
Air Traffic Controller
interface between the simulation software
accepts
and
processes
ATM/C
commands
statistics
(e.g.
responsible for
the
and
to
primary
follow, and
aircraft delay,
maintaining and updating
computer
simulation operating
etc.).
In
addition, it is
the traffic situation
Traffic Display in cooperation with the PDP-l1
It
displays ATM/C
sequence of runway operations,
for aircraft
average
as
and the ATM/C controller.
controller entries,
information such as the current
generated
functions
software.
on the
In the current
implementation the Command Activator also resides in this process.
The ATI/
Nominal
Plan
Flitght Plan Generator implements
Flight Plan
Generation.
Generation, Runway Scheduling,
It
determines
the
operations and the recommended flight
The flight
plans are then
by
the
optimal
and Traffic Flight
schedule
Command
process.
Activator
of
runway
plans to implement that schedule.
translated into ATC commands
Generator which resides in this
dispatched
three major functions:
by the Command
The commands are subsequently
to
the
ATM/C
controller
for
validation and subsequent transmission to the aircraft.
The choice
of functions performed by
each process closely matches
the real ATM/C environment where several systems operate in parallel and
communicate through various links.
sequential
require
machine
several
and to
accomplish
communicating
small intervals of CPU time to
parallel processing at least
Of course,
a computer is basically a
true parallel
computers.
processing would
Fortunately
by allocating
each process, it is possible to simulate
macroscopically.
Furthermore, the details
-64-
of how CPU
framework
for
CPU
can provide the
the Simulation Monitor
operating system while
the VAX
are the responsibility of
time is allocated to each process
adjusting
by
allocation
process
priorities
as
necessary.
The Simulation Monitor Process
3.4
simulation
creates
but
remaining
processes to
A.Lt
A-
*-.
not
It does
execution.
function,
J..
process controls and
Simulation Monitor
The
real-time
the
perform any
environment
operate properly.
"i
A
v
,JYe
A.a.~
monitors the overall
Air
Traffic Control
necessary
for the
It is also responsible for
A.LVW
t."CLL
V.
A.LIE
S.LJ.U1L.LVU_
monitor is shown in figure 3-3.
Data Base Initialization
3.4.1
We distinguish between two types of data in the simulation:
local data.
data and
and/or
modified) by
Global
data are shared (i.e.
all processes in
only be accessed by a single process.
is global.
the simulation.
global
can be referenced
Local data can
The major part of TASIM data base
Local data generally consist of working areas and other such
data.
Global data are categorized into 4 groups:
1. Process data
2. Dynamic data
-65-
Interprocess
Communication
Initialization
Terminal I/0
In it .aI iza t iorr-
Figure 3-3.
Top Level Flow-Chart of the Simulation Monitor
Process.
3.
Static data
4.
Screen data
Process data contain information required by the Simulation Monitor
to control and
by each
its operation.
process to control
number,
identification
statistics,
utilization
name,
process
of
number
former include process
The
execution
include
latter
The
etc.
information such as current simulation
process),
well as data required
monitor the various processes, as
process
CPU
specific
time (for the Simulation Monitor
and
active aircraft
currently
priorities,
total
number of
aircraft (for the Position Generator process), etc.
consist of aircraft "files"
Dynamic data
aircraft file
and radar "files".
one active aircraft in
contains pertinent information on
the simulation.
Four types of information are maintained:
include the aircraft type, flight
1. Aircraft ID data, which
transponder code,
number,
and other
such data typically
include
position, altitude,
found in a flight strip.
state
2. Aircraft
speed,
heading
data, which
bank
the
angle, and
other
such information
instantaneous dynamic
describing
the
aircraft.
Actual values
Each
state
of the
as well ask values indicated by
the various airborne or ground instruments are maintained.
-67-
data, which include
3. Aircraft command
and other such
speed* heading
the
data describing clearances
to receive
is due
received or
aircraft has
commanded altitude,
in the
future.
4. Aircraft scheduling
current sequence
data, which include
number, the assigned runway, the scheduled time of arrival
as well as other such
at the runway,
schedule
runway
optimal
under
data describing the
current
the
'
objective
function.
Each radar
file contains surveillance information
surveillance radar in the simulated
airspace.
be simulated though typically only one
provided by one
Any number of radars can
is used as the primary source or
position information at any given time.
Static
data consist
simulation environment
of reference information
which describes the
and remains unchanged throughout
the run.
group includes data describing the simulated airspace (i.e.
This
location of
airports, runway description, location and type of navigational aids and
radars,
departure
nominal
and
approach
routes,
etc.),
the
characteristics of simulated aircraft types, the traffic levels and mix,
etc.
change
Note that
with
even though the traffic level§ and
time,
determined apriori.
they
are considered
static
the traffic mix may
data
since
they are
-68-
Screen data consist of
on
CTID screen.
the
information specially formatted for display
This
group contains
base.
elsewhere in the simulation data
is in a
Here, however, this information
format suitable for quick display.
Even though this procedure
adopted since it conserves
computer memory. it was
requires additional
already found
information
required to repeatedly reformat the same
CPU time which would have been
and in a real-time environment
information whenever needed for display,
trading-off memory for CPU time is advantageous.
3.4.2
Interprocess Communication
is the "central nervous
Interprocess communication
simulation
very
design of
and consequently,
important
for
proper
system" of the
protocol is
the communication
operation as
well
as
for
maintaining a
flexible and expandable system.
Interprocess communication is accomplished
called
mailboxes.
process?
process
mailbox (receiver)
requesting for
mail,
A
performs the
sender of the
a message
(sender) leaves
providing vital
some function to
via virtual I/0 devices
information and (usually)
be performed.
requested function(s) and
in another
The
receiver reads its
(optionally) advises the
outcome and/or completion of the
operation.
The message
minimally contains the message code and the identification of the sender
process.
when
base.
It may
contain additional information
such information
is not readily
regarding the request
accessible from
the global data
For example. when some function has to be performed on a specific
-69-
aircraft,
whereas
can be retrieved
be required
that may
additional information
the message,
in
is included
identification
aircraft
the
from the global data base by the receiving process.
communications allow both asynchronous
This method of interprocess
is very
simulating
in
important
the
ATM/C
where some
environment
should be deferred
immediate processing while others
functions require
Tnis
to be easily implementable.
and sequential processing of messages
either because some higher priority function takes precedence or because
certain
successfully
be
to
have
conditions
before
function
the
can
be
asynchronous processing is
good example of
A
processed.
met
the message sent by the Simulation Monitor to the Air Traffic Controller
process
notifying it
availability of
of the
to the PDP-l1
Traffic Display Driver (i.e.
operation and prompt display of
be sent
data to
computer).
to the
To insure smooth
the latest information on the positions
of aircraft in the system, this function has to take precedence over for
example,
entered
on the
update has been
manually
of input
the processing
entered
keyboard.
completed.
or
that the
function is resumed
The latter
In contrast, when the
has validated
a
command to
Position Generator process receives a message
information in the appropriate
ot the request is waived by the
aircraft
state update
function of the two.
ATM/C controller
may have
when the TD
ATM/C controller has
some
aircraft,
the
in order to enter the new
aircraft's command data.
The processing
Position Generator until the end of the
cycle since the
latter is the highest priority
-70-
Subordinate Processes and Execution Control
3.4.3
The simulation
preparing
TASIM.
loading into memory and
monitor is responsible for
for execution
all the
subordinate processes
which comprise
Currently these are the Position Generator, the ATM/C Controller
process, and
the Flight Plan
Generator.
the Process data (described in the
The information
contained in
previous section) govern the loading
of all processes.
The functions performed
broad
categories,
triggered
by
intervals.
Monitor
periodic
the internal
All
which
by the simulation can be
and
aperiodic.
clock
and are
periodic functions
has
sole
are
divided into two
Periodic
performed at
controlled by
responsibility for
timing
functions
are
specific time
the Simulation
the
software and
the Simulation Monitor
controls the
achieving real-time performance.
The logic
execution
specific
of
according to which
periodic
functions
function performed.
is uniform
This is achieved
and
independent
of the
by assigning processes
into "timer chains" which define the sequence in which each process will
be activated in response to a specific timer alert.
there
is an
explicit
sequence of
functions
(or
Within each process
possibly
a single
function) which is performed in response to the timer alert message sent
by the Simulation Monitor.
Upon completion of the requested function(s)
each process sends a termination message to the Simulation Monitor which
causes the latter to look for and activate the next process in the timer
-71When the last process has been activated,
chain.
the expiration of the
waits for
and perform
time the timer
time interval at which
part of more than one timer chain
A process may be
chain is restarted.
the Simulation Monitor
different functions depending
on which timer
chain caused
the process activation.
Aperiodic functions are triggered by some simulation event wnich is
not
guaranteed to
generally
regularity.
performance of
The
involve the Simulation
any specific
occur at
such
time or
functions does
Monitor but is triggered by
with any
not necessarily
a direct message by
the process which detected the event to the process which is responsible
for
performing the
function in question.
command data triggered by a manually
The update
of the aircraft
entered command, is a good example
of an aperiodic function performed by the Position Generator process due
to an event detected by the Air Traffic Controller process.
The execution logic is thus
layered into several functional levels
with each level controlling the execution of the level immediately below
and controlled by the level immediately above it.
performed require
functions
that
triggered
their
minimal awareness
activation.
This allows
Within each level the
of the
chain of events
easy
modification of
function at any level by changing the sequence of subfunctions activated
and/or
inserting new
functions to
perform
as
functions in
be included in more
parts of
more
that chain.
Furthermore it allows
than one level in
than one
higher
this structure and
level functions.
In the
-72-
current
version
the
has a
simulation
single
timer chain.
periodic
Currently the periodic functions of TASIM are:
1. Traffic Generation
2. Aircraft Dynamics update
3. Air Data system update
4. Aircraft Navigation system update
5. Surveillance system update
6. Aircraft Status update
7. Command Activator
8. Traffic Display data update
functions are
first six
The
process while the
configuration
the Position Generator
the Air Traffic Controller
last two are performed by
In the current
process.
performed by
these functions
are performed
every 4 seconds which corresponds to the typical period of revolution of
the
in the
radar antenna
however
since a
there is enough
terminal area.
time chain can
CPU
time
for
This need
be executed
all
at any period
the required
performed.
Currently the aperiodic functions of TASIM are:
1. Runway Scheduling
2. Keyboard Input Editing and Processing
3. CTID update
not be
the case
as long as
computations
to be
-73-
process,
two
the other
while
by
is performed
scheduling
Runway
by the
are performed
Generator
Plan
Flight
the
ATM/C controller
process.
Initialization of Terminal I/0
3.4.4
and
computer
standard
the general
Since
system
operating
purpose terminal
was not
input, etc.),
of
a
by the VAX-ll/780
the
special functions
special purpose
driver was
designed and
designed specifically to suit the
The terminal driver was
needs of the
interface between the ATM/C controller
be discussed in
and movement.
driver provided
implemented.
and will
the proper
direct cursor control, item selection,
required by the simulation (e.g.
editing
cursor position
to perform
adequate
well as
information and initializing the
control the
required to
data
internal
screen in
the video
responses, setting
for display of tabular
operating mode
prompts as
includes formatting of
It
operation.
output
for real-time input
prepares the terminal
Terminal initialization
connection with the
and the simulation
ATM/C Controller process
(section 3.6.6).
When the initialization phase
process waits
real-time
initiates
the
Upon entering
Position
Subsequently the wait
number
the operator requesting
for input from
operation.
of occurrences
is completed, the Simulation Monitor
the
real-time mode,
Generator-ATM/C
state is entered again.
require
the beginning of
the Simulation
Controller
At
the program
timer
loop.
this point however a
Monitor's attention and
-74described in the
are
These
state.
wait
exit the
it to
will cause
following sections.
Timer Loop Alert
3.4.5
The timer loop alert indicates that it is time to start a new cycle
single timer loop is defined currently
Even though only a
loop alerts.
processing of timer
Figure 3-4 shows the
periodic timer loop.
of the
the logic has been designed to handle any number of timer loops.
The loop is first identified and if the last
has completed
execution the procedure
the delivery of a new alert at
for
first process
new timer loop
the end of the time interval appropriate
a START
question, (ii) sending
loop in
the timer
for initiating a
(i) requesting (from the operating system)
This includes:
takes place.
process in the chain
loop, (iii) collecting CPU
in the timer
message to the
usage and other
If
statistics from this timer loop, and (iv) reentering the wait state.
the
last process
loop has not
in the timer
initiation of the new cycle must be delayed.
occur -during normal
operation cannot
delay
in
requesting
interval
operations
be achieved due
the initialization
a
delay timer
(currently set
since they
Such situations should not
indicate
that
to CPU unavailability.
new cycle
of a
alert
completed processing, the
at the
at 1/100 of
end
The required
is implemented
of a
a second),
much
real time
by:
(i)
shorter time
(ii) initializing the
compilation of delay statistics, and (iii) re-entering the wait state.
-75
Figure 3-4.
Timer Loop Alert Processing Detail.
-76-
Delay Timer Alert
3.4.6
tr
The adea
that the initialization of a timer
-indica
glr
This situation is of
delayed due to CPU unavailability.
loop has been
course undesirable and is not expected to occur during normal operation.
has been a high number of other users (e.g.
15-20) on the system.
is well
usage
current CPU
the
that
the system
which
not incurred delays unless there
operates currently, the simulation has
indicates
in
environment
time sharing
the
in
even
Indeed,
below
This
the VAX-1l/780
capabilities.
The
figure
The
3-5.
timer
delay timer
the
logic for
processing
the
caused
loop that
alert is depicted in
alert
is identified,
statistics are updated to account for the additional delay interval, and
if the last
for
this
process in the loop has completed
timer
loop
is
initiated.
execution, the new cycle
Otherwise
the
timer
loop
initialization is further delayed.
3.4.7
Screen Update Timer Alert
This alert occurs periodically and causes the update ot information
on. the VT-100
superseded.
video
screen
This alert
ATM/C Controller process.
if
is common
the
displayed
to the
information
Simulation Monitor
Since it is an integral
has been
and the
part of the display
interface, its discussion will be presented in connection with the ATM/C
Controller process (section 3.6.61.
-77-
YES
Figure 3-5.
Delay Timer Alert Processing Detail.
-78-
Mailbox Message Processor
3.4.8
a message
When
The
sending process.
1.
the message.
received and
message
type of
the
depends on
processing
and respond to
order to process
wait state in
exits the
Simulation Monitor
mailbox the
in its
is placed
on the
The following message types are currently implemented:
READY message
message can
This
inform
to
loop
timer
a d
.
1
3
s
bee
nd
l a
c
--
-, e
t-10
process in the timer loop, if any, can be STARTed.
sending
process is the last
one in
processing is required since the
when
initialized
occurs.
a timer
that
Monitor
message (see sections
via the START
processing requested
3
Simulation
the
a periodic
process in
by any
be sent
the timer
next
If the
loop, no
timer loop cycle will be
alert or
a delay
timer alert
If the process is not the last one in the loop, a
START message is sent to the next process. -The wait state
is reentered in either case.
2. EXIT message
This
message is received by the
the sending
the normal
indicates
process is about to
Simulation Monitor when
end execution.
real-time mode, reception of
that
the
process has
While in
the EXIT message
encountered
some error
-79condition and
of the run is
thus an abnormal termination
system is already in
the termination
a reply to
an EXIT message
imminent.
If the
mode, this
message is simply
sent by the Simulation Monitor to all processes as part of
rundown
the
Simulation
condition
the
sent
the
message
It then enters
code.
messages are
EXIT
outputs
Monitor
that
process
In
procedures.
the
normal
the
of
the
identification
well
as
the
on-line and the wait state is entered.
as
the error
termination mode.
all processes
sent to
mode,
that are still
If an EXIT message
is received in the termination mode, the program checks if
all
exited.
If not,
all
processes
Monitor stores
the global
processes have
EXIT
message.
Simulation
When
all mailboxes, deassigns all
it waits
have
for a new
exited,
the
data base, deletes
I/0 channels and exits.
The
processing is shown schematically in figure 3-6.
3.4.9
Terminal Input Processing
When input
wait
state to
from the terminal
parse the input
is available, the
and perform the
process exits the
desired function.
commands are currently implemented:
1.
START
This
command
simulation.
initiates
the real-time
Optionally, the simulation
operation
of the
time at which the
Two
-80-
Figure 3-6.
EXIT Message Processing Detail.
-81-
that
the simulation
case
can be
should begin
operation
real-time
At that
pass through the "transient
up
building
if
steady
the
point the
This mode permits TASIM
real-time operation is initiated.
to quickly
In
FREERUN mode
in the
operates
specified time is reached.
until the
specified.
state
state" of traffic
performance
is
of
interest.
2. END
This command
causes the process to
mode, send EXIT messages to
enter the termination
all other processes, and wait
for replies before exiting (see previous section).
3.5
The Position Generator Process
The
aircraft targets in
radars
and
the
process
Generator
Position
the simulated airspace as well
navigation
equipment.
subsystems:
1. Mailbox Message Processor
2. Traffic Generator
3. Aircraft Status Update
4.
Aircraft Dynamics model
5. Air Data System model
movements
simulates the
It
of the
as the surveillance
consists
of
seven major
-826. Navigation System model
7. Surveillance System model
The surveillance
system drives all the
other components since all
other functions are performed once every revolution of the radar antenna
(see
discussion
of
the
Simulation Monitor
process
in
the previous
section).
3.5.1
Mailbox Message Processor
The operation
messages received
responsible
functions.
of the Position
in the
Generator process is
mailbox.
for interpreting
The mailbox
those messages and
message
controlled by
processor is
invoking the required
The messages defined in the current configuration are:
1. START
This message is sent by
a
new
invoked
update cycle
the Simulation Monitor and causes
to be
sequentially.
When
started.
All
the update
subsystems are
is completes a
ready message is sent to the Simulation Monitor's mailbox.
2. READY
This message is sent by
to
inform
the Position
the Flight Plan Generator process
Generator that
processing (i.e.
runway scheduling in the current configuration and traffic
plan
generation in
the future)
has been
completed.
It
-83allows the
Position Generator to
coordinate its requests
for rescheduling (see next section).
3. EXIT
This message
the end of
Simulation Monitor to signal
is sent by the
the run.
It causes the
position generator to
pause execution and to respond with an EXIT message before
exiting.
3.5.2
Traffic Generation Model
The traffic generation model is designed to realistically duplicate
in
detail the
processes which generate
Traffic is generated due to demand
airports
within
simulation
terminal
each airport
traffic mixes.
generation
the
These
of landings
time
the ATM/C
the terminal area.
for landing at, or
area
under
is characterized
and of takeoffs
taking off from
consideration.
by demand
are separate for landings and
Poisson random processes.
earliest
traffic in
rates as
In
the
well as
for takeoffs.
The
are assumed
to be incependent
The time of generation is
assumed to be the
controller has
knowledge of
some aircraft's
intentions to land at or takeoff from the airport
For
each airport,
an aircraft
(takeoff pr
landing) is generated
during each simulation update interval &t with probability
= X.At
p.
1.
1
-84at an airport i
demand rate for this type of operation
where A. is the
(number of aircraft per unit time)
it is well known that
the resulting generation times
will form a
Poisson process provided that At is small enough compared to the inverse
an aircraft, the
generation of
When this process
results in the
current simulation time
is assigned as
rate, 1/X. EDRA 67].
the demand
of
the aircraft's generation time.
The
aircraft
to
applicable
different for that
of takeoffs.
landing
for
The mix
airport.
each
the
on
based
is determined
type
Variation in the mix
aircraft
aircraft
mix
may be
over time may be
modelled as a step function or as a piecewise linear function.
are
Fixes
described
simulation
by
of aircraft that are can be
longitude, the types
the relative frequency
airport use
the
in
their
assigned
with which aircraft taking off
the fix, etc.
In addition each fix
latitude and
to the fix.
or landing at an
has several altitudes
that can be assigned to aircraft using it.
In order to determine the fix for the newly generated aircraft, the
model first determines
the fix must be
which
has just
been
compiled,
aircraft
all
eligible fixes.
In order
active and able to be assigned to
been generated.
the relative
operating
at the
Once
the liAt of
frequencies
airport
eligible,
the type of aircraft
eligible
with which
in question
to be
are
fixes has
each is used by
normalized.
The
-85assigned fix
is then randomly
selected from the
resulting probability
distribution.
In some cases the interval between the generation of two successive
landings
bound
for
the
same entry
fix
will
requirement for horizontal separations will
the traffic generation model simulates
taken
by
the
separations.
enroute
Let n
specific entry fix.
controller
desirable and the highest is the
the time t. of
that
be violated.
the ATC
In such cases
the actions that might have been
in
order
be the number of available
The model assumes
be such
to
provide
adequate
altitude levels at some
that lowest altitude is the most
least desirable.
For each altitude i,
the last aircraft passage is stored.
At
..
any time t the
altitude is called .Dp.en if
t. + t
< t
I
min-
where
t.
is
the minimum
time separation
between two co-altitudinal
aircraft.
The lowest
open altitude at
the time the aircraft
arrive at the entry fix is assigned.
time, the
earliest
time of arrival
time an
at the fix
altitude is open.
model then, is that the enroute
the
entry
aircraft
fix and
are
separations.
delays
handed-off to
is expected to
If no open altitude exists at that
is revised to
The
coincide with the
inherent assumption
in this
controller is aware of the situation at
incoming traffic
the terminal
in
order to
insure that
area controller
with proper
-86-
Flight Plan
Generator process is alerted through a
incorporate
the
functions.
Currently
new
in
aircraft
this
a cycle, the
been generated during
one aircraft has
If at least
decision
automated
the
support
Scheduling
Runway
the
triggers
message
mailbox message to
function which will determine the new runway schedule.
Aircraft Status Update and Handoff Processing
3.5.3
function
This
aircraft as
of all
status
is responsible
"life" in the simulation.
for maintaining
they pass
and
updating the
phases of their
through various
At any time the status of any aircraft is
1. Dormant
2. Active
Inactive
3.
4. Marked for deletion
generation all
Upon
by all
ignored
aircraft
full participant in
they
deactivated.
their
reached for
is
time
becomes a
assigned
update function
status
activation
until
functions of the
reach
their
final
Dormant
dormant.
aircraft are
exception of the
simulation with the
and the
Runway Scheduler.
aircraft,
some
aircraft are
is
Aircraft
the simulation.
destination
it
at
which
When the
activated and
remain active
time
they are
Landings become inactive shortly before touchdown on their
runway while
assigned
departing traffic
terminal
area
exit
fix.
is deactivated
Upon
upon reaching
deactivation
of
an
-87-
the now useless information.
(local) data bases discarding
purge their
When
this is completed, the aircraft
time
statistics are
a chance to
be given
subsystems have to
various simulation
aircraft,
is marked for
deletion at which
are saved if
global data
collected, the aircraft
needed, and the aircraft is finally deleted from the system.
it is appropriate
controllers that
and
to
automatic handoff
direction
start a
target ATM/C
mode,
acceptance
the
maintained until the handoff
of
aircraft determines
an
The ATM/C
and alerted through ATM/C
controller has
handoff
initialization and
handoff procedure.
are involved are identified
If the
messages.
handoff
position
The
activation.
whether
also handles
function
This
status
disabled the automatic
for
is explicitly accepted.
the
aircraft
is
Reminder messages
are sent if a prespecified time interval is elapsed without reception of
handoff acceptance.
3.5.4
Aircraft Dynamics
This
is responsible
function
for maintaining
and
updating the
dynamic state of all aircraft in the simulation, moving them through the
simulated region in response to
internal and external forces.
Internal
forces are generated to bring the measured aircraft state into alignment
with the commanded
variations.
magnitude.
state and include power ch'anges
External
forces include the effects
The aircraft model
relationships
between
and control surface
of wind direction and
used accurately describes tne functional
various aircraft
components, and
includes time
-88-
lags present in the longitudinal,
aircraft.
lateral and vertical response of each
The model was implemented by Heinz and is described in detail
in EHEI 76] and [HOF 72J.
3.5.5
Air Data System Update
This
function
manages
airborne instruments
airspeed/mach
indicator.
models
neinz
updates
readings
such as the compass/gyro
indicators,
the
The measurement errors for
developed
by Hoffman
available
(heading indicator), the
altimeter,
and
vertical speed
all the instruments are based on
[HOF 72].
The
model was
function is responsible for maintaining
airborne
simulating
navigational equipment
the
navigation
system
for
implemented by
different
devices
Omnidirectional Range
station
to the
are
errors.
modelled:
By
and
and (ii)
aircraft and
large,
aircraft
radio navigational aids.
(i)
Very
(VOR). which indicates the
aircraft,
current readings of
each simulated
navigation today is performed with the use of
the
the
Aircraft Navigation System
This
Two
of
LnsI 7.6.
3.5.6
the
number
and
High
Frequency
magnetic bearing from
Distance
Measuring equipment
(DME-TACAN), which indicates the slant range between the station and the
aircraft.
As is generally the case, the simulation assumes that VOR and
DME-TACAN transmitters are collocated in VOR/DNE or VORTAC installations
so that
both the range and
the bearing from a
single known geographic
-89location
is available.
Each
tracking
two
VORTAC
navigation
VOR/DME or
modes
to be
radial interception
aircraft
stations.
simulated
3.5.7
This
while tracking a different
of independently
allows a variety of
navigation
VOR course.
and VOR
The models
and their errors were developed by
The model was implemented by Heinz EHEI 76J.
Surveillance System
This function
aircraft
models the radars
position information.
generally employed
skin tracking).
Two types of
transmits a coded
and
altitude
Control with
surveillance systems are
StUrveillance radars
aircraft.
ground transmitter (often referred to
Beacon trackers
aircraft which
sometimes
that provide Ground
in Lracking airbUrne
utilize reflected energy from the
as
capable
including area
used for aircraft navigation systems
Hoffman [HOF 72J.
is
interrogate a
transponder on the
reply, consisting of
information.
Normally
the
the aircraft ID
two
systems are
operated in parallel, with the surveillance radar serving as a backup to
the
beacon
system.
The simulated
radar
parameters which determine its performance.
for
those parameters
that are used
is described by
a
set of
Table 3-1 gives the values
to describe the
terminal area and
enroute versions of the current civil_ beacon system known as ATCRBS (Air
Traffic
Control Radar
Beacon System).
have not yet been implemented in TASIM.
Note that
tracking algorithms
-90-
TABLE 3-1
-
ATC Radar Beacon System Parameters
Parameters
Terminal
Enroute
Max range
60 nmiles
200 nmiles
Elevation coverage
0 to 45 degrees
0 to 45 degrees
Scan rate
15 RPM
6 RPM
Range bin width
1/16 nmiles
1/4 nmiles
Azimuth bin width
0.088 degrees
0.225 degrees
-91-
3.6
The ATM/C Controller Process
managing
the
controller
controller, and for
information display
tabular
(CTID)
and the
The traffic display software reside partly in the
traffic display (TD).
and partly
VAX-l1/780
the ATM/C
input by
real-time commands
processing
accepting and
for
is responsible
process
Controller
ATM/C
The
in the
It will be
PDP-11/34.
described here,
the ATM/C Controller process.
however, since it is an integral part of
activator is a function of the
In the current configuration the command
ATM/C Controller process.
3-7 presents
Figure
Controller
hardware
video
and
terminal
of the ATM/C
functional diagram
software components.
identifies the major
process and
(VT-100
a top level
calligraphic
MEGATEK
The
display
computer) are also depicted in the figure.
The
message in
mailbox message
the mailbox.
Management software
In
turn, it invokes the
The
Traffic Display
sends
appropriate
Data
sent
by
the
Input
START message is
Traffic
formats and
Display driver
updates the positions of the aircraft
(resident in the PDP-1l/34) which
on the MEGATEK screen.
the
of a
Traffic Display Data
Management software
data to
aircraft
the reception
Activator when a
and the Command
received.
the
processor is invoked by
The Traffic Display driver also responds to data
Processor
to
modify
the
(displayed range, displayed altitudes, etc.).
is triggered periodically
by a timer alert and
screen characteristics
The CTID Display software
updates the information
*Driver
managementl
fInput
MProcessor
.
S
Simulation
Base
Base
Mailbox Messages
Figure 3-7.
raf
Traffic
Disp>1ay
Data
Management
Mailbox
Message
--ProcessorACITO
VAX-11
PDP-11
DATA LINK
COMMAND
VTO
Functional Diagramn of the Air Traffic Controller Process.
-93-
on
controller inputs.
Each input character is processed separately as it
This allows various
is entered.
line by defining
the input
is triggered by
The Input Editor
the VT-100 video display screen.
editing functions to
special function keys.
be performed on
When
a line ending
character is entered, the input editor transfers the command line to the
for identifying and performing the
Input Processor which is responsible
requested
The Input
function.
Processor
displays warning
and error
controller in case of
messages on the VT-100 screen to inform the ATM/C
incorrect, ambiguous or unrecognized input.
3.6.1
Mailbox Message Processor
messages
Two
are
currently
recognized
by
the
Mailbox Message
Processor:
1. START
This message, sent by
ATM/C
controller
the Simulation Monitor, directs the
process to
initiate
aircraft positions on the radar
Command Activator function.
an update
of the
display and to invoKe the
When both these functions are
complete the process responds with a READY message sent to
the
Simulation
Monitor.
This currently
ends
the only
timer loop of the simulation (see section 3.4.5).
2. EXIT
This message, sent by
the Simulation Monitor, signals the
-94-
end
of the
the
I/0
run.
The ATM/C
channels,
mailboxes,
stops
controller process deassigns
associates
the
with
display
driver
PDP-ll/34 computer, responds with
the VT-100
software
and the
on
the
an EXIT message sent to
the simulation monitor, and exits.
3.6.2
Command Activation
This function is responsible for maintenance and timely dispatching
of commands to the aircraft.
Piloting commands are generated either by
the Command Generator based on the
input by
the ATM/C controller
automatically are
in the form
format as the ones entered by
aircraft flight plan or are directly
For uniformity the
commands generated
of alphanumeric strings and
the ATM/C controller.
in the same
Processing of both
types of commands is therefore identical.
Two concepts for command management
ATM/C systems there
is no precise future planning so
issuing a piloting command at some
anticipated.
Even
can be identified.
In present
that the need for
future point in time cannot be fully
when, for instance, the
ATM/C controller knows that
soon some aircraft will be required to make a turn to intercept the ILS,
he does
which
not have the means
this command
to determine the exact
should be executed.
time or location at
Furthermore it is not at all
clear that the pilot will accept a command to turn to some heading, say,
90
seconds
from now
consequence the ATM/C
when
he is busy
preparing for
controller waits and issues the
landing.
As a
command when the
-95-
"queueing the commands on
refer to this process as
We
it immediately.
pilot has to execute
certain point in space and the
aircraft reaches a
the ground".
We
in the air".
"queueing commands
opposite
the
visualize
contrast,
by
can,
Here commands are
process,
namely
transmitted to the
aircraft in advance and it is the pilot's responsibility to execute them
at
the
time.
proper
between
distinguish
We
location-triggered commands, depending on
starts at a specific point in time
time-triggered
and
whether the command execution
or at a specific point in space.
An
IFR flight plan is an example of queueing location-triggered commands in
the
air
approved
process
since it ~is the
terminal
the
where
followed by
of~the~ pilot to
the pilot
The
area environment.
is
of this
only time-triggered
execution of a holding
perform one
responsible to
minute turns
In
advanced ATM/C systems
dimensional flight planning,
time-triggered command
1,2 or 3 minute straight legs.
incorporating four
follow the
procedure is an example
today's ATM/C system is the
command sequence in
pattern
approach
A missed
route.
in
responsibility
queueing in the air is certainly feasible (from a technological point of
view) when coupled
with a digital air-to-ground data
link and airborne
time-triggered autopilots.
The
simulation
was designed
with the
capability to
time-triggered and location-triggered commands.
handle both
This is accomplished by
associating four different time values with each command:
-96-
1. the scheduled time of transmission
2. the actual time of transmission
3.
the scheduled time of execution
4.
the actual time of execution
these depends on the method employed
The interpretation of each of
by
the
communication
time and
commands
are queued
and scheduled)
is zero as
times in that case measures the
the
the actual
piloL workload.
When
the difference
in transmission times
long as the
command is transmitted
discrepancy in execution
The
executed.
it is scheduled to be
before
of
execution is the pilot
time of
associated with
in the air,
ground the
congestion
difference between
the
the actual
Lherefore
is
response
(actual
contrast,
transmission and
of
time
By
links.
and/or
workload
controller
ATM/C
by
the
the actual time of transmission is
difference between the scheduled and
caused
queued on
are
commands
When
system.
ATM/C
capability of the pilot and/or aircraft
autopilot to execute the commands promptly.
There are
even though this is feasible
most
compelling
compatible
of
(to the
all
the air
for advanced automated ATM/C systems.
is the
need
develop
to
greatest extent possible)
Queueing commands on the ground
ground (namely
queueing commands in
distinct disadvantages in
systems
The
that are
with existing equipment.
only requires advanced equipment on the
the computers and
computer programs that
implement the
advanced ATM/C functions) and is capable of operating with absolutely no
advanced
capability
requirements
onboard
the
aircraft.
Second, by
-97commands to
transmitting
the changing traffic
congestion
create
in advance
on
the
we either, (i)
that have not yet
the capability to revise commands
deprive the system
been executed if
the aircraft well
situation warrants it,
link
communications
by
or (ii)
transmitting
we
and
subsequently revising or cancelling commands.
The
ground
ATM/C
system concept
thus avoiding
all these
described here
problems.
queues commands
Aircraft
on the
flight plans, and
therefore commands, are subject to change until some short time interval
before they are scheduled to start execution.
sent
to the
to inform him
ATm/C controller
transmitted to the aircraft shortly and
not done so yet.
to review the
validation,
This interval
At that time a message is
that the
to request validation if
he has
is chosen to allow the ATM/C controller
traffic situation and possibly modify
thus
command will be
providing another
check against
the command before
system malfunction.
When the command has been validated
by the ATM/C controller the command
is transmitted
to the aircraft via
data link, if such
proper time of
execution or the software provides
ATM/C controller so that he
is used, at the
display cues for the
can correctly time the command transmission
by voice.
3.6.3
Traffic Display Driver
This
function
is responsible
displayed on the TD screen.
for
maintaining
the
information
The displayed information and capabilities
of current ARTS III displays was chosen as a basis for the design of the
-98-
The
Display.
Traffic
simulation's
goal however
primary
not to
Instead the aim
spectrum of ARTS III capabilities.
duplicate the full
was
that will easily allow incorporation of
was to create an infrastructure
new functions and visual aids that are needed by the ATM/C controller in
order
perform
to
his
in
the
MEGATEK display screen.
ground
automated
control
a photograph of the
is a blown-up negative of
Figure 3-8
environment.
role
new
The displayed items include:
1. Aircraft symbol and tag
The
same
symbol,
configuration,
system.
MATM/
ATM/C
line
a
is used
(\) while
aircraft
n
teL
(/) for
of
the
aircraft in the
contrl
oVf
thef=u
a backward slanting
around
The orientation
be
the
ATM/C
orientations are available.
included in the aircraft tag:
Flight
identification
other
appropriate
the
aircraft
is positioned relative to
aircraft symbol.
a.
current
easy identification.
is centered
(surveillance) position
by
the
uncontrolled traffic is represented by a
position
selected
in
direct
represented by
slanting line
aircraft tag
line
to represent all
ircra
supervisor are
forward
slanted
The
estimated
target.
The
the center of the
of the aircraft tag can
controller.
Eight possible
The following information is
i
code (transponder
identification
pilot-controller communications.
code or
used
in
50,11-56
Figure
3-7.
Traffic Display Layout
-100-
b.
Indicated altitude (in 100's of feet)
c. Latest altitude clearance (in 100's of feet)
(+ if the
symbol
speed
d. Vertical
aircraft
is
climbing and - if it is descending.)
e. Ground speed (in knots)
In
a
addition
this
If the
number.
etc.)
is in
aircraft
some
unusual
state,
hazard or conformance
hand-off/hand-over,
however, (e.g.
alert,
computer identification
the aircraft
area displays
(five
Normally
tag.
the aircraft
is provided on
characters)
area
message
purpose
general
an appropriate blinking mesgpge is
displayed
in this area.
2.
Terminal area network
The terminal area network consists of "nodes" and a set of
arcs
connecting
It defines
them.
air routes
airspace.
Each type of node
symbol.
Two
intersection
be
position
of
(waypoint),
obstructions
ground
also
types
represented
displayed terminal area
within the
departure
is represented by a distinct
nodes,
are
is available.
accompanied by their name
nodes
if
VORTACs
and
VORTACs
currently
(hills, radio
as
arrival and
nominal
defined.
antennas,
their
and
airway
etc.)
Fixed
can
existence and
waypoints
are
code while for obstructions the
minimum safe altitude is displayed.
-1013. Range rings
Three concentric range rings
They
whose
are
TCA
centered
around
are includea in the display.
the location
is simulated.
They
are
of the airfield
used to provide the
controller with a measure of distances.
4.
Airfield layout
A rough sketch depicting the
airfield as well as other
runway layout of the primary
airfields in the simulated area
is displayed.
5. Simulation clock
The current simulation time in hours, minutes, and seconds
is displayed.
Each
depending
item
its
displayed with
minimum
on the
relative
screen
is displayed
importance.
maximum intensity while
beam intensity.
ditferent intensiLy
Aircraft symbols
range rings are
A medium setting
clock, the terminal area network
with
is used
and
tags
are
displayed with
for the simulation
and the airfield layouts.
In addition
to the standard display items, special purpose displays can be overlayed
on the
screen.
Figure 3-8
shows such a special
purpose display.
operation of this display is presented in chaptr 4.
The
-102Input Editor
3.6.4
The editing
and
of
display
character
basis.
function is responsible for
ATM/C controller
Three
types of
manging the accu mulatio
keyboard
inputs on
keys are
defined.
a
character by
Self-insert keys
transmit printable character codes that are inserted in the input buffer
and
on
displayed
display
Instead they cause simple
Control keya
not stored
are passed immediately to the
processing of normal
in the input
key commands and
input identification function for further
o f quick acinky
rcesn
Th
transmit
editing functions to be performed on
Finally Quick action keys are single
the input line.
prcsig
screen.
are generally
characters which
non-printable
buffer.
the
is
input lines which allows some
independent of the
actions to be taken
without destroying the normal input already in the input buffer.
3.6.5
Input Processor
The input
the
processor is the
simulation facility.
available for
parsing.
heart of the
It is triggered
The input
man-machine interface of
whenever a
command line is
processor is responsible for parsing
the command line, and identifying and performing the requested function.
Parsing
linguistic
the normal
sense
input lines
of the
word
resemble* syntax
and is
usually
analysis in the
more difficult
if
the
command language is flexible enough to provide a convenient and reliable
tool
of communication
between the
ATM/C controller
and the software.
-103-
currently
The
the
has
parser
implemented
features which
following
enhance its usability:
1. Abbreviation ot verbs and keywords
2. Default settings for required but missing keywords
3. Implicit specification of missing parameters
4. Checks for input consistency
first
The
and
controller
ATM/C
probability of error.
as
system
Abbreviating verbs
having totally
the
and keywords for example make
different meaning.
understood by the
cases however,
In most
among them.
that are inconsistent
result in inputs
erroneous commands
increase
also
they
command will be
that a misspelled
it more probable
while
software
the
between tie
of communications
the ease
three enhance
As an example, a clearance for an aircraft to climb to 10,000 feet while
something is wrong, though it is not
it is now at 20,000 indicates that
by
any
means clear
addressed.
descend
Probably
the wrong
possible
that the
ATM/C
It is however
instead
information
what is wrong.
on
of
climb
that
or
current
the aircraft's
checks for data consistency. have
vast
and
deserves extensive
the
study
altitude.
in future
controller meant
controller
ATM/C
been included.
aircraft was
A
variety
has wrong
of such
The subject is however
research
regarding the
design of the man-machine interface of the automated ATM/C system.
-104-
CTID Update
3.6.6
CTID
The
function in
update
the modified data
Flagging of
modified.
Information Display has been
Controller Tabular
on the
data
after some
triggered
function that performed
is the responsibility of the
the modification.
screen
to provide
the ATM/C controller
system
status and
performance.
significant departure
display
been
has
from current practice
the
that
display is a
a secondary
TD to also
which uses the
against two separate
The argument
ATM/C
data on the
with alphanumeric
use of
The
supplementary information.
displays
the VT-100 video
Information Display uses
Controller Tabular
The
controller
at
look
can
the
supplementary data without being distracted
from his primary task, that
of monitoring of target movements on the TD.
Minimizing the distraction
is of course desirable.
from the ATM/C controller's primary task
It is
not clear however that use of a secondary display will result in greater
distraction if
it is directly adjacent
emphasis has been
time little
to the TD screen.
given to the
At the same
distraction resulting from
the additional effort required to access supplementary information.
second display will
visible at
thus minimize accessing time
current
CTID
configuration
divided
into
four
sections
The
following
information.
displayed:
more data can be
provide more display area so that
one time and
is shown
each
types
in
figure 3-9.
displaying
of
a
The
and effort.
The
different
information
are
The
screen is
type
of
currently
-
X-. -6,
L- PD TYPE
|CtRE
CAV
A
-EA
T
___
-ere
iaq: L.50 mriutes
-FRE7RD ELAY SHIFT
: ET9 I
~
*...
Naim dela: 4.75 minutes
ivage shifts: 2
.
-Cumt
eb.iective
Ai-
AASLi-3
AAS4
: MINItZE AVERACE DELAY
CUidi T2 j CU
TR~
Mt
S±
A
-.Runwe in use... 2
Parvatter setting,- 29.95
Uind fron 200 at 10 knots.Ceiling at 3000 feet:
-.Visibilit'a 3 NM .
7
-
r
TL.TO IANJU DIRECT; SPEED 200; DESCENID TO 60 - E0:08:00
'iL LY 40 FROM HTM; SPEED 180; DESCEND TO 50
09:30
TR HDC 90
SPEED IM0 DESCENT TO 40
90 13:00
Figure 3-9.
Controller Tabular Information Display Layout
-1061. Airport scheduling data
2. Simulation statistics
3. ATC messages
4. Computer generated commands
When there
is more data
of some type
screen section on which the data
scroll
the
defined
for
different
section
data within
this
the
purpose.
screen section.
however.
than there is space in the
is displayed, the ATM/C controller can
display section
Currently
each
Several types
In that case
using quick
type of
of data
the ATM/C
action keys
data
occupies a
can share
controller will
a screen
be able to
chose which data type he wants displayed at any one time.
3.7
Implementation Status
TASIM
is currently
fully
operational
conventional terminal area ATM/C system.
for
simulations
of
The only conventional function
which has not yet been implemented is tracking of radar targets.
ATM/C automation
software,
has
already been
Plan
Generators,
Monitor
software
runway scheduling is
implemented.
of
the
Development
of the
Command Generator,
is, therefore, required
undergone
approximately
60
performance has been satisfactory.
hours
of
Of the
the only function that
and
before
Nominal and Flight
of
the Conformance
experimentation and
testing of the fully automated ATM/C system can begin.
far
the
debugging
TASIM
runs
has
and
thus
its
To date, however, there has been no
systematic testing of all its functions.
-107-
CHAPTER 4
THE RUNWAY SCHEDULING PROBLEM
4.1
Introduction
The
runway
scheduling problem
addresses the
Given a certain configuration consisting
of Nr runways, and the current
position ot a
set of Na aircraft wishing to
each aircraft
i, a runway n
runway, STAR.
criterion,
spectrum
land or takeoff, assign to
and the scheduled time
of arrival at that
such that some objective function, or system efficiency1
,
is optimized.
of operational
there can
following question:
The
schedules are
and safety
constrained
requirements.
be substantial improvements
As
by
a wide
Dear has .shown,
in runway capacity
and aircraft
delays when runway scheduling is applied. [DEA 76]
Scheduling
variation
in
improves runway
the
minimum
standards between pairs
time
efficiency by taking
interval
of aircraft.
type
etc.).
of operations
Methods
operations
have
involved
for
determining
been
developed
by
ATC separation
These intervals are
ATC safety requirements, and vary depending
the
allowed
advantage of the
based on the
on the type of aircraft and
(arrival-arrival, arrival-departure,
the minimum
in
connection
time
with
intervals between
runway capacity
(1) The term efficienc.y is used to stress the fact that runway capacity
maximization is not always the main objective.
Runway capacity can
be traded off against reductions in aircraft delay, as well as
against more equitable distribution of delays or other user costs.
-108-
Appendix A
([PAR 81a], [HOC 74]).
models.
illustrates how scheduling
affects aircraft delays through an example involving aircraft of varying
landing speeds.
A substantial improvement can also be realized through coordination
of
basis.
current practicedepartures
In
stream allow it.
the landing
some
for both landings
scheduling decisions
due to
critical length.
departure
queue
dissipates.
soluion to^
which seems
adequate gaps
unavailability of
prbl
implementable,
queue increases beyond
If the departure
is the
As
, thi
+pr-ioa
meho
is
inevitable today can
in the
request until the
with
case
necessarily efficient.
but not
when gaps in
are inserted only
are created by tower controller
landing stream, gaps
on a tactical
and takeoffs
desgne
all
A to
be
U
easily
--.
runway time
Much idle
be eliminated through
procedural
scheduling the
usage of the runway system.
Runway scheduling
is tactical.
The schedules
the known
traffic at
any given time.
schedule will
have to be
updated every time
on
intention to land or takeoff known to
known aircraft, the
are generated based
This means
that the runway
a new aircraft
the ATM/C system.
possible formulations of the problem
makes its
Given a set of
fall into two
distinct categories:
The static formulation assumes that every aircraft in the system is
capable of arriving at the runway
threshold (for takeoff or landing) at
t .
Stated differently. this implies that
or after some reference time
-109-
the current position of aircraft in
the terminal area (or the gates) is
of no importance to the scheduling problem.
as
long
as they
exception
of
constrained
assumption
the
t .
greater than
aircraft chosen
only
the
by
simplifies
in
time
which
accounted
for
to
ininimum
greatly
formulations
implicitly
are
the
is not
through
All STAR's will be feasible
As a
consequence,
operate first,
separation
problem
an
STAR's are
requirements.
since
explicit
minimum
the
with the
time
This
it allows
some
variable,
but
intervals
between
is
consecutive operations.
The
dynamic
broad framework of
above assumption
airspace or at
aircraft
addresses runway
ATM1/C in the terminal area.
is weak since
As
the
earliest
a result
each
scheduling
the position (within
time
that
within the
It recognizes that the
the aprons) and the operational
affect
threshold.
formulation
the terminal area
characteristics of each
it can
aircraft's STAR
reach
the
runway
is constrained
by the
position of the aircraft at the time rescheduling occurs.
4.2
Definitions
This section gives formal definitions
are
extensively used
here are the
for terms and variables that
throughout the chapter.
ones applicable only to specific
defined as they are needed.
Variables not appearing
t
sections.
Those will be
-110We will generally assume that there are Na aircraft to be scheduled
When aircraft are
on Nr runways.
classified into types, Nt will denote
the number of distinct aircraft types.
Lower case
i and
Usually
letters i,
j will
j, k,
1, m, and n will
to specific
refer
be used as indices.
aircraft, m and
n will index
runways, and k and 1 will be used to index aircraft types.
The set of aircraft for which
is called the decision aircraft ae,
A
d
scheduling decisions have to be made
A d, and is defined as:
= {l,2,...,N
}
The set of active runways is denoted by R and defined as:
R ={,2,...,N r }
Two functions, which provide
aircraft and those
the
denotes
the correspondence between indices of
of runways and aircraft types,
index of
the
runway assigned
are defined.
to aircraft
RWY(i)
i. Similarly
TYPEi) denotes the index of the aircraft type to which i belongs.
The
schedule
of runway
operations will
be denoted
by S
defined as the set
S = { STAR. : i=1,2,...,N ,
a
in
n=1,2,...,N
r
}-
4-1
and is
-111STAR. is the time aircraft i
in
where, as before,
will be set to
if n/RWY(i) and it
n. STAR.in is meaningless
at runway
is scheduled to arrive
zero by convention.
formulation of
In the mathematical
integer 0-1
through
specified
will be
assignment of
and the
of operations
the sequcnce
the runway scheduling
variables
problem
aircraft to runways
e..
Ij nm
These
.
are
defined as follows:
if n = RWY(i), and m = RWYj)
1
and STAR.
in
ij nm
jm
jand
if i
e..
< STAR.
P
= m = RY(i)
4-2
0 otherwise
i and
different aircraft
for two
i.e.
j,
if e..
=1,
aircraft i is
scheduled to operate before aircraft j.
given
a
Under
set
of
separation
and
interval allowed between two aircraft i
m respectively will be denoted by s.
standards,
.
minimum time
the
j assigned
to runways n and
The subscripts used stress the
dependence of the minimum time separations on the aircraft pair i and js
as well
two
as the runway on
operations
not have
do
for
requirement
which each aircraft operates.
s..
to
be
to
be
in direct
Note that the
The only
succession.
applicable is that aircraft
i precedes
Ij nm
j,
is set to
By convention, s..
e.. =1.
ijnm
ijn
negative number when runways n and m are independent.
aircraft
i.e.
a large
-112-
that
imply
scheduled
cannot
be
chapter 2, section 2.2.2,
time constraints, discussed in
The lead
time
revised.
Ad
aircraft set, A.
are not
aircraft
Such
or their
the runway
at
of arrival
A , whose
by
aircraft, denoted
of
a set
be
there will
runway assignment
the decision
included in
have to be taken into account in
However, they still
the new schedule since they affect the schedules of other aircraft which
The runway schedule, therefore, is
are still eligible for rescheduling.
Ad, while aircraft in tne set A0
optimized over all aircraft in the set
appear
in
maintained.
the
The
constraints
set A+
that
insure
to
aircraft in
of all
separations
proper
the terminal
are
area ATM/C
system can now be defined as:
A=
+ 0o
d
A UA
Latest Feasible Time of Arrival at
The concept of the Earliest and
Runway
the
was
definitions by making
as the aircraft.
formalize their
explicit their dependence on the
runway, as well
chapter
in
We will denote them by
depend
LFTAR.
in
and
now
introduced
on
the
2.
We
in and LFTAR.in .
EFTAR.
characteristics
aircraft
EFTAR.
in
(primarily its
airspeed) and represent the time interval within which eacn aircraft can
reach
the
runway threshold
from
its current
position.
For landing
aircraft, this interval reflects the flexibility provided by the TFPG in
expediting
the
aircraft through
delays imposed by the schedule.
the
terminal area
and
in absorbing
Note that, if holding is allowed at the
entry fixes, LFTAR. for landing aircraft
in
will not be limited by flight
planning considerations before the aircraft reach the entry fix or while
-113-
they are holding there. At that stage of the approach, fuel availability
limiting factor for LFTAR. .
onboard the aircraft is the
in
For takeoffs.
.
EFTAR. reflects the taxi time to runway n, while LFTAR.in will typically
in
be
very
large since
there need
be any
not
how long
restriction to
aircraft may be on gate hold.
that, given the proper
When LFTAR. is very large, it is possible
in
conditions, aircraft i will be pushed at the end
time
schedule
runway
the
possibility,
an aircraft
the position
In
is revised.
of
. POS. and
min
aircraft i
max
i
avoid
to
in the
in
the overall
such a
sequence of
We will denote by minPOS
and
i can occupy
and maximum positions that aircraft
in the sequence within a certain runway n.
position
order
can occupy
operations will need to be constrained.
POS. , the minimum
in
max
of the sequence every
The limits applicable to the
sequence
will be
denoted by
POS..
i
Finally, in the mixed integer linear programming formulation of the
next section, we
wil.1 use the symbol M to
denote a very large positive
constant.
4.3
Formulation of the Dynamic Runway Scheduling Problem
We
are
now
in
the
position
to
formlate
the
dynamic runway
scheduling problem as a mixed integer mathematical program.
Find
STAR. and
in
, for i,jEA
e..
ijnm
and n,mER,
that minimize some
specified objective function Z(S) and satisfy the following constraints.
-114-
1.
Scheduling Constraints
jm
M
jimm
V
.V
b)
2.
STAR .
jcAd
-"LER
< LFTAR. 'e ..
3m
jm 33mm
Sequencing Constraints
V jcA d
e..
i.
a)
E
icA+
>
>-
.
POS. *e..
n
jm 3jm
V meR
ifj
b)
e..
<
POS. 'e..
Zjmm - max
Jm JJmm
x
-iPA
V j6Ad
V mER
i#j
c)
-
e.
> minPOS
V jeAd
e ..
<
V jEAd
neR mcR iEA
i#j
d)
neR mER iEA
POS.
-115-
3.
Spacing Constraints
+ s..
ijnm
ijnm 'e..
ju --> STAR.in
STAR.
-
(1
-
e..)'
)jnm
V icA+ and jEA+
V nER and mER
4.
Assignment Constraints
S{0,1}
e..
1j ff
b)
V iEA , jEA
V nER , mER
V
e..
jeA
d
mER
cE
(ij
)=
+ e..
j imnf
nm
1
V jEA
d
+
V iE-A
-{j}
nER mER
ifj
d)
z
e..jriijm- < M'e.jm
jjm
V jEA , V mER
nER iEA
ifj
-
e)
< M'e..
e..
j imn JJaM
d
V jEA , V mER
nER iEAI
The scheduling constraints (la and lb) simply require that, for all
aircraft
j
j,
STAR.
jm
lies in
the interval [EFTAR. ,LFTAR.J
isassigned to runway mi (i.e.
if e..*
J 1m
when aircraft
j n
= 1). When aircraft
j
is not
-116to be zero, which
assigned to runway m, the two constraints force STAR.
is consistent with the convention adopted for that case.
and 2b,
the summation of
aircraft
e..
assigned to
e..
over all
ijmm
operate on
i and
is zero unless both
j.
scheduled to operate before
In constraints 2a
operate similarly.
sequencing constraints
The
aircraft i, is
runway m
j
When
prior to
aircraft
j,since
runway m and
are assigned to
j
the number of
i is
is not assigned to runway m, the
right hand side of the constraint becomes zero.
Consequently, all e..
1JMM
are constrained to be zero.
and
2b except
that they
Constraints
apply to
2c and 2d are equivalent to 2a
the position
of aircraft
j in
the
overall sequence.
In
the spacing
compared to the
to 1, i.e.
aircraft
that
problem variables and parameters.
when aircraft i
j operates
STAR.
constraints, M is assumed to be
is greater
than STAR.
in
jin
minimum time separation between the
e..
When
is assigned to operate
on its assigned
runway m,
by at
a large constant
e..
.5isequal
on runway n before
the constraint insures
least s.
in
two operations.
,
the required
In all other cases
is zero and the constraint is redundant, since the right hand side
is dominated by M.
Finally, the
sequencing
assignment constraints insure that
variables
sequencing variables to
are
consistent.
Constraint
the two integer values 0
the values of the
4a
and 1.
limits
the
Constraint 4b
requires that every aircraft is assigned to a runway once and only once.
-117Constraint 4c insures that either aircraft i is scheduled after aircraft
j, or j is scheduled after i, but not both.
4e state that if e..
jjMm
all
aircraft
i and
is zero,
runways
e..
ijnm
n.
Finally, constraints 4d and
and e..
This
jimn
insures
should also be zero for
that each
aircraft is
consistently assigned to one runway.
The
objective
function of
function
the schedule S
examines runway
in
the above
(i.e.
formulation
of the STAR. 's).
in
efficiency measures and
is a general
The next section
develops alternative objective
functions to be optimized by runway scheduling.
4.4
Measures of Runway Efficiency
Runway
efficiency
is
a
multi-dimensional
quantity.
It
is
impossible to capture every aspect of the operation of the runway system
in a
single number.
collectively,
We,
describe the
therefore, use several
term.
In
measures which, taken
this work
we will
consider the
following measures of runway efficiency:
1. Runway capacity
2. Aircraft delay (total or average)
3. Weighted aircraft delay (total or average)
4, Equitable distribution of delay or fuel costs
The
mathematical
several sections for
expressions
which
are
each of these measures, are
developed
in
the next
functions of the time
-118-
each aircraft
to these expressions
We will refer
reaches the runway.
interchangeably as cost functions or objectiye function.
Runway Capacity
4.4.1
Runway capacity is the most
of the
normal use
The
runway system.
term refers
saturation capacity of the
to the
Saturation capacity is defined as the average number of
aircraft that can
runway system during some
takeoff and/or land at the
time, assuming infinite
unit of
widely used runway efficiency measure.
models have been
A number of
demand.
developed to estimate analytically the saturation capacity for a variety
Of runway
operations
coufigurLatius.
to
operations, At.
calculate
The
the
mean
Lime iLervals
Lhe
These modeis use
interval
time
bVeLween
between successive
saturation capacity is then simply
the inverse of
At, i.e.,
At
is averaged
sequence,
over all
and over
4-3
= 1/A t
C
successive pairs
all possible
of aircraft
The
sequences.
sequence is weighted by its probability
for a specific
contribution of each
of occurrence given some mix of
aircraft types.
The saturation capacity, as quantified
based
on
a
representative
aircraft
above, is an expected value
mix.
(1) See for example [PAR 81a] and [HOC 74].
It is,
therefore
an
-119We are interested in optimizing
inappropriate measure for our purposes.
at a particular instant in time, i.e.
the capacity of the runway system
based
on
takeoff.
This
the capacity
wishing to
traffic
of the
"snapshot"
a specific
capacity will generally be
substantially different from
Naturally, in the absence of
calculated in equation 4-3.
by
lead time constraints or other operational restrictions,
optimizing
situation changes, we
as the traffic
the "snapshot capacity repeatedly
land and
also optimize the long term capacity of the runway system.
The average interval between successive
Consider some schedule S.
is given by the difference between
operations achieved by this schedule
the
scheduled
operation,
time
the
of
by
divided
the
last
total
operation
number
of
and
that
of
operations,
the first
i.e.
the
cardinality of the set A+
At(S) = (max {STAR.
isA+
in
4-4
neR
nER
The runway system
min {STAR. }) / MA+)
-
in
icA+
throughput rate can now be defined
as the inverse of
the average interval between successive operations:
TR(S) = 1 / 6t(S)
Throughput rate is
maximized if
last operation is minimized.
can
thus
be
used
4-5
as the
the interval between
the first and the
The last (or maximum) scheduled time (MST)
objective
function
Formally the objective function can be written as:
for
the optimization.
-120-
Z(S) = MST(S)
= max { STAR.
}
. eA+in
4-6
nER
and
section 4.3
optimization of
the
is a
minimax problem
since the
objective becomes:
min Z(S) = min
idA+
max { STAR. i
. . + +in
iEA
neR
nER
function can
new objective
The
normal minimization
new decisinu
min
by a standard
variable t a,
converted back
be easily
transformation, i.e.
by
into a
defining a
by substituting
t
max
for the above objective and by adding the following set of constraints:
V jEA , V mER
STAR. < t
jm - max
4.4.2
Aircraft Delays
Delays
system
are
experienced
responsible for
direct operating cost
the
by
increase in
aircraft
a
within the
large fraction
of airline
(fuel, crew salaries, elIc.)
the block time
cost is also increased due to
for the trip.
terminal
is
area ATH/C
costs.
The
increased due to
The indirect operating
the lower aircraft utilization rate (e.g.
passenger miles per unit of time).
The cost of the trip as perceived by
-121-
is also increased
the passenger
air travel.
average speed of
decrease the
since delays effectively
These effects on
airlines and passengers
alike are particularly obvious on short trips.
experienced by an
The delay
aircraft in the terminal
the aircraft arrives at the assigned
difference between the actual time
the time it would
runway and
The latter
traffic.
runway, PTAR.,
have arrived in the
is called
We note that the
absence of other
area is the
the preferred
absence of any other
time of
arrival at the
runway assigned to the aircraft in the
traffic will be the one the
aircraft can safely reach
So if we let PTAR.in be the prefe.rred time of arrival at runway n
first.
we have:
PTAR.
2
=.
min d { PTAR.ini
iEA
nER
For planning purposes we will use STAR
instead of the actual time
of arrival at the runway which, of course, is not known.
The aircraft delay is given by:
d. =
STAR.
- PTAR.
nER
or
d.
= S TAR
WY(i) - PTAR.
since STAR. is non-zero only for n = RWY(i).
in
Several delay related objectiv-e functions can be of interest.
-122-
is simply given
aircraft delay, TAD(S),
The total
all the aircraft in the
delays experienced by
system.
by summing the
For some runway
schedule S therefore,
N
a
d.
TAD(S)
i=1
The average aircraft delay is given by:
AD(S)
= TAD(S) / N
any
For
constant.
instance
a
of
the
runway
scheduling
problem.
Na
is a
The total and average aircraft delay are therefore equivalent
objectives, since the schedule that minimizes one will also minimize the
Furthermore since
other.
N
a
TAD(S)
=
(
~1
STAR
iRWY(i)i
- PTAR.)
i=1
Na
Na
STAR. RWY()-
i=1
and
de lay
the second
is.
i-1
term is a constant.
minimization af
equivalent .tao minimiz ing the. Aum
The objective functions considered
the cost associated
4-7
PTAR.
average aircrafl
f .the. STJAR'Ia.
thus far implicitly assume that
with aircraft delays is the
same for all aircraft.
-123This
delay
weighted
the unit time of delay
importance attached to
aircraft to vary.
total
The
true.
generally
is not
allows the
experienced by different
We define
N
a
w.d.
TWD(S)
1 3.
i=1
importance,
the relative
w. is
where
or weight,
unit time of
of the
delay experienced by aircraft i.
will
weights
The
typically
associated with a unit time of
same weight will be
and the
be
representative
delay (e.g.
of
the
costs
the direct operating cost),
applied to all aircraft
of the same type.
i.e.
w. = w.
1
J
if TYPE(i) = TYPE(j)
This assumption is necessary for
chapter.
In general
some algorithms presented in this
however, each aircraft may be
weight.
The average weighted delay is defined as:
N
AWD(S) = TWD(S) /
w
i.=1
)
assigned a distinct
-124Since the denominator is again a
constant for each instance of the
problem, the two objective
runway scheduling
functions are equivalent.
Similarly, the weighted sum of the STAR's,
N
N
WSST(S)
w. (
i=1
N
STAR.
)
n=1
a
=
4-8
w.STARiR(i)
i:l
is also equivalent
to TWD(S)
and to
AWD(S) and
will be
the general
objective function used when discussing optimization of aircraft delays.
We now turn to an even more general case of delay related objective
functions, where the contribution of each aircraft is additive, but each
contribution is not necessarily a linear function of the delay.
We will
refer to this class of objective as Generalized Weighted Delays.
N
a
4-9
f.(d.)
GWD(S)
i=l
Note that f. may be distinct
a
special case
of GWD(S)
for which f.'s
delays, i.e.
f.
:1
=w.d.
i 1
for each aircraft, and that TWD(S) is
+ (constant)
are linear
functions of the
-125-
User Cost Distribution: The CPS Methodology
4.4.3
of delay (or other cost) distribution
We finally turn to the issue
the
among
Inequities
system.
in the
of
of usersI
classes
users, or
distribution of
area ATM/C
the terminal
almost always
user costs
of multi-user systems is performed.
arise when system-wide optimization
Such inequities will not be tolerated
by the users, and it is essential
that they are dealt with by the optimization process.
Objective functions can be developed
the
cost (instead
maximum user
optimization criterion,
of the
for example,
to remedy this situation.
the optimum
as the
schedule will be such
that all user costs will tend to be roughly the same.
this approach is that such
is used
average cost)
If
The drawback with
optimization criteria are generally too weak
to result in significant improvements on a system-wide basis.
A
preferable
method
which
insure the
schedules
users.
cost
impose
on
constraints
non-preferential
treatment of
the
runway
all system
A wide variety of explicit constraints can be imposed to achieve
this goal.
can
is to
For example, the average cost experienced by each user class
be constrained
experienced by
constraint
will
not to exceed
all the system
a certain percentage
users collectively.
alleviate large inequities
of the average
This type of
4ut there is no guarantee
One obvious
(1) Users may ,be classified in several different ways.
classification may be by aircraft -type, while in discussing cost
distribution, it may be more appropriate to put all aircraft of each
airline in a distinct user class.
-126-
that
class will
some user
consistently discriminated against.
not be
The reason is that they do not remove the cause of the inequities.
They
merely limit their effect.
is best maximized
efficiency
characteristics.
aircraft
runway scheduling problem
in the
Inequities
of the same
where those
maximum efficiency.
with similar
of aircraft
by "bunching"
in the case
is particularly evident
This
arise because runway
landing speed should
of landing
be bunched for
Departing aircraft behave similarly with respect to
the departure route assigned to them and to the runway on wnich they are
to take
As a
off.
consequence, "minority" aircraft,
those that
i.e.
pushed back in the schedule.
The phenomenon
avoided
of "bunching" suggests that
by restricting
rather than
the position
by restricting the
This
experiences.
aircraft in
amount of delay each
therefore,
limit the position
constraints which
The Constrained
work will,
of the
inequities can be best
group ot aircraft
concentrate on
of the aircraft
Position Shifting (CPS) methodology,
the sequence
a
class of
in the sequence.
proposed by Dear,
is designed to achieve this goal [DEA 76].
CPS
seeks
aircraft in
to
limit
the forward
the ATM/C system.
and
4ackward
The nominal sequence
movement
of all
of operations (as
(1) See section 4.6 for further discussion of this effect. Also
extensive analysis of aircraft
[DEA 76) and [PSA 78) include
bunching for the case of scheduling landings on a single runway.
-127basis
the
as
is used
PTAR's)
relative
the
by
determined
for the
constraints imposed on the position of aircraft in the optimal schedule.
The
from the
can
nominal position that
to any
arbitrarily set
be
is set to N
When MPS
sequence is unconstrained.
of zero
an MPS
For
and
is feasible
MPS
only the
optimization is
no
to N , better and
a
better (from the
function) schedules can be
of the value of the objective
point of view
determined.
value.
MPS increases from 0
As
achieved.
any aircraft.
will be allowed for
operations
of
sequence
nominal
the largest deviation
(MPS) is
of Position Shifts
Maximum number
of all aircraft in the
the position
Experience has shown that MPS values of 4 or
5 achieve values for the objective function that are very close to those
At the same time it insures
without any position shifting constraints.
non-preferential treatment of all aircraft in the ATM/C system.1
The
constraints
sequencing
CPS
implement
the
allowable
position
methodology.
in
for each
min POS in) are expressed in absolute terms.
max
Ci) + MipS
POS. = POS
in
nom
and
min
POS.
in
POS
(1) See for example
nom
[DEA
Ci)
-
MiS
76]and [PSA 78]
of
seccion 4.3
maximum
and minimum
formulation
Note that
sequence
in the
the
the
aircraft
One could let
maxPOS.
and
-128to make the relationship of the limits
on the position of aircraft i to
its position in the nominal sequence. POS nom(i)
explicit.
The absolute
form of the sequencing constraints allows the CPS methodology to be used
as
For
schedule.
some
example, we can
aircraft by
to the
aircraft
operate in
variety of
imposing a
a means of
setting the
have direct control of
we can force
direct succession if needed.
made available
for that
two aircraft to
Such options can
controller, as a means
to the ATM/C
to the
the position of
maximum positions
minimum and
Similarly,
same value.
operational constraints
and will be
of controlling the
scheduling function of the automation software.
4.5
The Complexity of the Runway Scheduling Problem
For each problem
of the quantity
we can identify a number
which is representative
of input variables defining it.
the size of the problem.
This number is called
size of the runway scheduling problem may
The
be the number of 0-1 variables in the formulation of section 4.3 i.e.,
SIZE(RSP) = (N N )2
a r
The time complexity,
of size
n, is defined as
time units
T(n), of an algorithm that solves some problem
some function f(n) such
required to obtain
a solution is
that:
the number of
equal to Cf(n),
for some
constant C. We then say that the time complexity of the algorithm is af
Iag order aL f(n), and write:
T(n) = O(f(n))
-129is a property of algorithms rather
Even though the time complexity
the time complexity of a
problem itself, we can talk about
than of the
mean the time complexity
problem if that is understood to
of the most
efficient algorithm that can solve the problem.
time complexity of the runway scheduling
In order to determine the
.
and
Consequently, the formulation is a mixed integer linear program
function is also a
long as the objective
(MILP) as
the
on the
of the problem variables, STARi
schedules are all linear functions
e..
that are imposed
set of constraints
note that the
problem, we
minimized.
This
variables.
problem
-
transformation
can be
the
is
MST(S)
When
obviously the
is
objective
obtain an
used to
linear function of
case when
function,
equivalent MILP
WSST(S) is
a
standard
as shown in
section 4.4.1.
In the case of generalized weighted delays there is no exact method
which will transform the problem into the standard MILP formulation.
can however
approximate this problem
function by
weight
a piece-wise
as an MILP if
We
we substitute each
linear approximation
(see [BRA 77]).
Furthermore, if all the generalized weights, f (d ) are convex functions
of
d.,
aircraft delays
the
the approximation
does not
increase the
complexity of the problem.
The
(e.g.
time complexity
Branch
O(exp[(N aNr ) 2]).
of the
and Bound) that
best known
general
solves the runway
urpose algorithm
scheduling problem is
This value can be improved by recognizing that, due to
-130-
the assignment constraints of section
is not independent.
aircraft
all min.
We will show
problem is exponential with respect
complexity of the
in the
's
For example. if e.inn is set to 1 for some value of
has to be zero for
n, then e..
4.3, the selection of the e..
system and
polynomial with
that the time
to the number of
respect to
the number of
active runways.
To make the discussion concrete, a simple algorithm that solves the
problem will be
analyzed.
It consists of generating
ot values for eij nm s and solving
such set.
Obviously
sets of values.
the remaining linear program for each
the algorithm will have to
generate only feasible
Generating all possible sets of
the assignment constraints
the feasible sets
values and then using
to reject the ones that
are infeasible will
not be an improvement over the classical branch and bound approach.
Feasible
sets
of values
can
be generated
efficiently
by first
values for e..
's, and
then generating all the possible sequences within each runway.
From the
assigning aircraft
to runways, i.e.
selecting
13-nn
aircraft sequences the rest of the values can be generated in polynomial
time.
The number of ways aircraft can be assigned to runways consistently
N
is (Nr) a since each aircraft can be assigned to any
runway (i.e. Nr
alternative
ways)
independently
of
any
other
aircraft
assignment.
(1) Note that, with the exception of the case where GWD(S) is the
objective function, the solution to the resulting LP is trivial.
-131-
Furthermore,
runway assignments,
specific set of
for a
the number of
distinct sequences that can be constructed are:
N
r
lK !
n=1
where. Kn is the number of aircraft assigned to
is of the order of N ! since
runway n.
This number
it is (at least in principle) possible to
assign all aircraft to one runway leaving all tne others idle.
The time complexity of the runway scheduling problem is therefore,
N
= 0(
T(RSP)
(N ) a N !
r
Instead of generating
a
)
all sequences within each runway
limit the number of distinct sets
the sequencing constraints to further
Suppose that each aircraft is only
of values that have to be evaluated.
allowed
to be
shifted by
nominal position.
we may use
(forward or
at most MPS
backward) from its
Let us consider the following recursive algorithm for
generating sequences within each runway:
Step 0: Determine the next aircraft in the nominal sequence that
has not been assigned a
position.
been assigned positions go
If all aircraft have
to step 3.
Otherwise remove
this aircraft from the list and continue with step 1.
Step 1: Scan the list ot
If the
allowable positions for this aircraft.
list is empty.
go to step la.
Otherwise go to
-132-
step lb.
Step la: Reset the list of allowable positions for this aircraft.
Reenter
the aircraft
position
assignment.
level
of
to the
list of
aircraft without
the
previous calling
Return to
the algorithm.
If this
is the
top calling
level then STOP.
Step lb: Assign to the aircraft a position from the list that has
not been already assigned to another aircraft and delete
the
position
the
from
list.
the allowable
If all
positions have already been assigned go to step la.
Step 2: Apply this algorithm on the aircraft that remain without
an assigned position.
Upon return from the next calling
level go to step 1.
Step 3: Generate
the e..
jm
objective function
's from the
sequence.
and return to the
Evaluate the
previous level of
the algorithm.
This
tree.
algorithm in
trave:sal of a
The nodes on the k h level of the tree represent all the possible
position assignments
tree
essence performs "depth-first
for
runway n
to (k-1) aircraft.
is Kn+1, where,
aircraft assigned to the runway.
The
Accordingly the depth
as before,
K
n
is the
of the
number of
number of branches from each node
on the kth level correspond to the number of allowable positions for the
k
th
aircraft.
(1) The term depth-first means that the branches out of any node are
traversed before any additional nodes on the same level are reached.
-133In order to determine the number of branches from each node, let us
assume that at the kth level we are considering the aircraft i for which
POSnom (i)=k.
From
the tree there will
the root of
be exactly (MPS+l)
branches since the first aircraft will be allowed to occupy positions 1,
The second aircraft will be allowed on positions 1,
2, ...
,
MPS, 14PS+1.
2, ...
,
MPS+1, MPS+2.
But since one of these positions will be occupied
by the first aircraft, there will only be (MPS+1) feasible positions for
this
aircraft
(K -MPS)
there
as well.
By the
same reasoning,
branches from each
will be (MPS+l)
number of branches from nodes at the
number of positions
at each
node.
level above
Finally,
the
last MPS levels are limited by the
that are not yet occupied as
opposed to the number
of allowable positions.
Consequently the total number of
the
total number
of feasible
leaves on the tree, and therefore
sequences that
need to
be evaluated is
given by:
(K -MPS)
(MPS)l (MPS+1) (Kn-M
Each
step of
Furthermore, each
the algorithm can, be performed
leaf of the
in polynomial time.
tree requires the recursive
algorithm to be performed at most
K
times.
part of the
The time complexity of the
algorithm is therefore dominated by the number of leaves that have to be
reached.
-134-
is when
worst case
the
Again
are assigned
all aircraft
to one
By considering MPS to be a constant the number of leaves is
runway.
N
0( (MPS+1) a
number of possible runway
combine this result with the
We can now
assignments to obtain:
N
T(RSP)
= O(
r
in terms
is polynomial
which
4-10
[N (IIPS+1)] a
of
the
number
runways,
of
Nr,
and
exponential in terms of the number of aircraft, Na
4.6
Variations of the Runway Scheduling Problem
In
this
section
exact
for which
problem
scheduling
examine
we
all
only
cases
three
variations
algorithms have
the
runway
appeared
in the
of
aircraft
are
considered.
of section
4.3
are assumed
landing
literature.
In
Furthermore,
the
non-binding.
According to the classification of section 4.1, therefore,
constraints
scheduling
the variations to be discussed here belong to the class of static runway
scheduling problems.
The
algorithms
Instead we
possible and
not valid.
will
will focus on
not
be
described
in
the assumptions that make
how the algorithm
The goal will be to:
fails when any of
quantitative
terms.
the approach taken
these assumptions is
-1351. Explore
the special
characteristics of
the problem that
emerge,
2. Identify
that may
the
desirable
properties of
be used to
general
optimal solutions
advantage in obtaining
problem, as
require additional
the
well
solutions to
as undesirable
constraints to be imposed
ones that
in order to
insure that the resulting schedules are implementable.
3. Gain insight on various aspects
of the problem and on how
their interactions affect its complexity.
4.6.1
Scheduling Landings on a Single Runway (MPS=infinity)
This is the simplest in
It was studied
the class of
runway scheduling problems.
extensively by Dear, [DEA 76].
In
our terminology1 the
problem can be stated as follows:
Given Na aircraft
wishing to land on a
single runway and assuming
that
1.
EFTAR. = t
PTAR.
3..
2. LFTAR. = o,
3.
4.
max
POS.
1
. POS.
min
1
o
.
V i=1.2,....N
a
V i=l,2,...,N
= N
aa,
= 0
(1) Runway subscripts are
suppressed.
V i=1,2,....N
V i=1,21...,N
a
not needed for this discussion
and have been
-136-
arrival at the
scheduled time of
schedule which minimizes the maximum
find the runway
runway (MST(S)).
is some reference time which
Here, t
can be arbitrarily set to zero.
Dear
that
proves
solution
the optimal
analytically derived, and it is unique.
speed and precedes all others of
this an
calls
Dear
speeds) sequence.
Dear
of landing
which each aircraft
as one in
a descending sequence
succeeds all others
lower (or equal) landing
higher (or equal) landing speed.
ascending (in terms
similarly defines
can be
The optimal schedule is the one
aircraft succeeds all others of
in which each
problem
to this
of higher landing speed and
precedes all others of
lower lainuLg speed.
Analytical
solutions
constraints are imposed
derived
also
are
if
initial
particular. it is
In
on the landing sequence.
final
and
assumed that there exists one aircraft with fixed landing time STAR =t ,
0 0
which is constrained to land
constrained to occupy the last position in the sequence.
scheduled
time
of arrival
if, which is
first and another aircraft,
at the
runway
for i
Of course, the
is not
fixed.
The
optimal solution in the constrained cases is shown to consist of at most
three
subsequences,
either one
ascending and
two descending,
or two
ascending and one descending.
The above results
the
matrix
of
time
are based on the special
separations between
results, therefore, cannot be extended
structure exhibited by
successive
landings.
These
to cases which include departing
-137aircraft because the arrival-departure separations
do not have the same
structure.
important aspects of
bring out several
Dear's results
the runway
scheduling problem.
characteristics
similar
with
aircraft
of "bunching" of
demonstrate the phenomenon
1. They clearly
the relationship
result of
is a direct
Furthermore, the bunching
in this case).
landing speeds
similar
(i.e.
between landing
speeds and minimum aircraft separations.
2.
initial
Imposing
schedules
the runway
solution
in
two
a
have a multiplicity of solutions
affect the value
does not
ways:
sequence, (i.e.
number of aircraft in each
since the actual
function.
on
a single ascending
unique solution), we now
subsequences
constraints
optimal
the
affects
instead of
First,
final
and
Second, aircraft bunching
since, even though aircraft of
of the three
of the objective
is not as pronounced
the same landing speed are
in direct succession within each subsequence, they can now
be
in" up
to
different
three
places
in
the
overall
sequence.
3. In every
terms
of
sequences.
which
case, the optimal
aircraft
schedules
Behind this
allows
a
solutions are not
unique
but
in
defined in
terms
aircraft
transformation lies an assumption
schedule to
be
derived
from a
-138sequence
of operations.
auLpion.
In this
We will call
case, the
it the scheduling
scheduling assumption is
simply that the runway will not remain idle unnecessarily.
Since no scheduling constraints are imposed, the scheduled
time of
arrival at the
runway for any aircraft
j can
be
determined by:
STAR. = STAR. + s..
where.
i
sequence.
13
1
J
is'the
The
algorithms
subsections.
aircraft
directly preceding
j
in the
same scheduling assumption is used in the
which
will
be
examined
in
the
next
two
A more general version of this assumption is
used in the heuristic algorithm described in section 4.8.
4.6.2
Scheduling Landings on a Single Runway (MPS<N -1)
This version of the runway scheduling problem is similar to the one
that was described in the
previous subsection.
Now however, sequencing
constraints are imposed on some or all aircraft.
The algorithm does not
depend on any particular structure of the sequencing constraints.
it was
that
developed based on
each aircraft
i can
the CPS methodology however,
be shifted
up to
problem can be formulated as follows:
we will assume
MPS positions
backward from its (unique) nominal position, POS
nom
Since
forward or
(i). Accordingly the
-139Given Na aircraft wishing to land
on a single runway. and assuming
that,
EFTAR. = t
PTAR.
1.
1
1
=c2,
2. LFTAR.
3.
4.
,
0
V i=1,2...,N
V i=1,2,...,N
3..*a
POS. = POS nom (i)
I
max
= POS nom (i)
. POS.
min .
2
,.**.N
+ MPS ,
V i=1,
-
V i=1, 2 ,.*.,Na
iPS ,
runway schedule which
IPS are constants, find the
where. POS nom(i) and
a
minimizes the maximum scheduled time of arrival at the runway (MST(S)).
Psaraftis. [PSA 78],
this
solve
First
problem.
programming algorithm to
developed a dynamic
that,
showed
he
the
given
scheduling
assumption presented in the previous section and ignoring the sequencing
constraints, the
Salesman Problem, (TSP).
solving
TSP's
constraints.
This
for
function go to
as a classical Travelling
problem could be formulated
Furthermore,
be modified
could
was
the dynamic programming approach
letting
done by
the sequencing
incorporate
to
of
the value
the objective
was reached.
infinity whenever an infeasible state
The
approach can also be modified to handle the total aircraft delay and the
weighted sum of the aircraft
delays as objectives.
The time complexity
of this algorithm can be shown to be:
N
T
At
this point
reduced the
(RSP) = O(Na 2 a)
Psaraftis made
time complexity of
a key
assumption which drastically
the algorithm from an
exponential to a
-140-
He took advantage of the
polynomial function of the number of aircraft.
fact
into categories
can be classified
that aircraft
with similar characteristics.
(or types) each
These types can be defined such that, for
all i=1,2,...N
a
s..j
TYPE(j)
= TYPEk) iff (
=
each category,
within
all
5 1}
= s.
s..
i.e.
S.k
and
aircraft are
indistinguishable with
respect to their time separations.
This
original
assumption
state-stage
formulation.
The
allows
a
more
compact
diagram associaLed
worst case
(in
representation
the
dynamic programming
with the
terms of
of
time complexity)
for the
modified formulation occurs when each of the aircraft types contains the
same number of aircraft.
Letting Nt be the number of aircraft types and
denoting by rxl the smallest integer which is greater or equal to x, the
time complexity of the modified algorithm is
N
TMDP (RSP) = O( Nt EN a/Ntt1+1]t )
The time complexity of the
function of
the number of
modified algorithm is then a polynomial
aircraft.
It remains
with respect to the number of aircraft types.
T
(RSP) are
the same type.
identical when N t=N a
i.e.
exponential. however,
As expected. TDP(RSP) and
when no two
aircraft are of
-141Unlike
the
analytical
programming approach
results
derived
does not depend
aircraft
separations.
validity
of
Instead,
the optimality
by
Dear,
on a particular
the only
recursion
the
structure of the
condition necessary
is that
the
dynamic
for the
separation matrix
satisfies the triangle inequality, i.e.
S..
+
ij
for
all
aircraft
satisfied,
last
triplets
S.
jk
> S.
ik
i, js and
k.
When
this
the information required at each stage is
scheduled
aircraft as
well
as the
number
inequality is
limited to only the
of aircraft
in each
aircraft class that have not yet been scheduled.
We
can generalize
inequality
is false
this result by
but,
for
any
observing that
if the triangle
i, j, k,
four aircraft
1, the
inequality:
s
+ sjk+ skl > sii
is satisfied,
we can assure correct
on
two scheduled
the
program.
last
Similarly,
aircraft
spacing by maintaining information
at
every-stage
if the equivaLent inequality
of
the dynamic
is satisfied for all
aircraft n-tuples the state representation of the dynamic program has to
maintain information on the last n-2 aircraft in order to assure correct
separations among
all aircraft pairs.
In thik general
complexity of the algorithm is:
T
(RSP) = O( Nt(n-2) [[N a/Nt 1+1] t
case, the time
-142-
This
landings are
to be scheduled.
not
if
hold
j
case, however,
the
is not
inequality.
may
pairs
between two
of n for mixed operations is 5 or
is. it is possible to insert
2 or 3
departures between some
the required separations
aircraft without increasing
of landing
triangle inequality
aircraft- scheduled
departing
is a
takeoffs and
when both
In particular. the
landings. i and k. The typical value
6. That
satisfies the triangle
aircraft the separation matrix
For landing
between them.
It
constraints are
handled.
Namely, each
conditions
that
instead of
the same way the
The
scheduled time of
now the
sequencing constraints are
state would now have to
one.
satisfy two feasibility
requirement would be
only additional
arrival at
it may seem
At first,
studied by Psaraftis.
can be handled
that these
scheduling
programming formulation to the
included in the dynamic
runway scheduling problem
if
happens
what
consider
to
is instructive
well as the
the runway as
value of the objective function associated with each state would have to
be
stored.
this
Unfortunately,
is the
case
only
when
tne runway
throughtput rate is maximized.
When other
to guarantee
the dynamic
perspective,
objective functions are optimized.
an optimal solution
program is not
when
this approach fails
because the optimality
satisfied.
criterion for
Looking at it from a different
scheduling constraints
are imposed
the state-stage
description is incomplete unless the time variable is introducea as part
-143-
of it. In particular, time has to
be the stage variable along with the
number ot aircraft for which scheduling decisions have been made.
more
Another,
subtle,
the
is that
longer valid.
types is no
aircraft into
problem
In the
classification
of
absence of scheduling
constraints, it is reasonable to assume that aircraft of a specific type
in their relative
will land
order in
which aircraft of
objective function.
are
In
present.
nominal order.
a certain class
This is no
In any
land has no
effect on the
longer true when scheduling constraints
essence, aircraft
each type
within
each has distinct limits
indistinguishable since
case, the relative
are
no longer
on its scheduled time
of arrival at the runway.
4.6.3
Scheduling Landings on Independent Runways (MPS<Na-1)
We now
are active.
turn to the
In
case where two (or
more) independent
all other respects the problem
runways
considered here is very
similar to the ones considered in the previous two sections.
Given N
aircraft
wishing to land on two
independent runways, and
assuming that,
(1) Two runways are independent when aircraft operations on the two
In effect, each runway can be
runways need not be coordinated.
scheduled as if the other did not exist.
-1441.
PTAR.
2.
LFTAR.
EFTAR. = t .
o
in
in
00
in
V i=1,2,...,
V n=1,2
V
,
3. maxPOS.in = POS nom (i) + MPS
4.
. POS.
in
min
=
POS
nom
i) -
V i=1,2,...N a
V n=12
S
V i=1,2,-...,N
a
V n=1.2 '
,
runway schedule wnich
MPS are constants, find the
where, POS nom(i) and
a
V n=1,2
minimizes the maximum scheduled time of arrival at the runway (MST(S)).
In the absence of any sequencing constraints, this problem could be
to
extended
a
as
formulated
dynamic
solve it
program
and
This
[PSA 78].
Psaraftis'
algorithm
approach fails
can
be
when sequencing
constraints are present.
author developed
The
Bound algorithm
were incorporated
by introducing
to solve this
was used and sequencing
The 2-tour TSP formulation
problem, [PAR 78].
constraints
a Branch and
artificial nodes
in its
graph representation.
In the graph representation of the unconstrained problem, each node
an aircraft
represents
aircraft was
possible
represented by a
position
constrained
problem can
modified graph.
larger
than
and
that
number of nodes each
assignment
runway
be shown
to be
for
a 2-tour
nodes on the new
The number of
of the
modified graph, each
In the
to be scheduled.
graph
of the
corresponding to a
the
aircraft.
TSP defined
The
on the
graph is considerably
unconstrained
problem.
Both
-145problems, however, are shown to be
constrained
airft
problem
of the same time complexity, i.e the
is exponential
with
respect
to
number .0
the
in the system as opposed to the number of nodes in the modified
graph.
In addition
has the
to having exponential time
same problems as
complexity, this algorithm
Psaraftis' dynamic programming
approach with
respect to scheduling constraints and inclusion ot departing aircraft.
4.7
Heuristic Versus Exact Algorithms
The three variations of runway scheduling discussed in the previous
section
provide good
evidence of the
complexity of
the problem.
The
most striking observation is the rapid increase in the complexity of the
problem as
new constraints are
section 4.6.1
are
gained
in the
aircraft into types
in
section
The unconstrained
can be solved analytically.
introduced (section
advantage
imposed.
Finally
all
becomes exponential.
programming approach
is quickly lost when a
4.6.3.
Once sequencing constraints
4.6.2), the problem
dynamic
problem of
The
by classifying
second runway is introduced
approaches
fail
when
scheduling
constraints or departures are introduced.
The
complex
general runway
scheduling problem
than
the
Furthermore, a
any
of
variations
new solution has to
(landing or departing)
(&ection 4.3)
that
have
be found every time
enters the terminal area ATM/C
is far more
been
examined.
a new aircraft
system.
In view
-146of the restrictions imposed on the algorithm by the environment in which
it has
to operate, it is necessary to abandon the
search for strictly
optimal schedules.
The alternative is to use heuristic algorithms which
will generate near
optimal schedules within the time
limits imposed by
the real time operation of the ATM/C system.
Heuristic
algorithms consist
generate new feasible
way.
The
of a set
solutions to the problem at
rules are usually local
solution that has
of rules that
in nature.
function, to
generate a new
therefore, a
local variation of
hand in a systematic
They are
thus far resulted in the best
feasible solution.
the current best
are used to
applied to the
value of the objective
Each
new solution is,
solution.
The newly
generated solution is compared to the current best and the one producing
the
best
terminates
value
when
for the
the
objective
function is kept.
current solution
The algorithm
is better than
all
its local
variations generated by the algorithm.
There are
First,
two disadvantages associated
the solution
found by
with heuristic algorithms.
the heuristic
may be
far from optimal.
Second,; since the heuristic rules are usually local in nature, the value
of the final solution may vary greatly depending on the initial solution
used
to
start
performance of
may,
the
algorithm.
In
many
a heuristic algorithm can
situations,
the
be ascertained.
for example, have a "worst case performance of 2".
worst case
An algorithm
This means that
-147the
value
of
the
objective
function
resulting
from
the heuristic
solution will be at most twice that of the optimal value.
The
average
loss
in
performance with
respect
to
the strictly
optimal solution is usually much less severe than indicated by the worst
case performance
based on
of the algorithm.
pathological cases which
Worst case performance
is usually
are seldom, if ever, encountered in
real applications.
A
much
performance,
more indicative
i.e.
on the average.
2 may,
optimal.
measure
how close to
for a
heuristic is
the average
the optimal are the heuristic solutions
A heuristic, for example with worst case performance of
on the average,
generate solutions that
are within 10%
of the
An extreme example of this discrepancy between the average and
worst case performance of algorithms is the Simplex method used to solve
Linear Programs.2
In the worst
case, the time required
optimal solution using the Simplex
to obtain the
algorithm is an exponential function
of the number of constraints.3 The average performance of the algorithm,
however, is a polynomial function
of this number.
Simplex method is, understandably, due
average
versus
worst
case
The
success of the
to this tremendous difference in
performance.
Unfortunately,
the average
(1) Assuming the objective function is to be *Minimized. The analogous
definition applies for maximization of the objective function.
(2) In this case the performance is in terms of time required to obtain
the solution since the Simplex method is an exact algorithm. The
principle, however, remains the same.
(3) See [PAP 82]
-148performance of an
algorithm is often much more
difficult to derive (or
even estimate) than its worst case performance.
The
heuristic algorithm
important
advantages which
this application.
First, the
which has. been implemented
make it a very
example.
that
process of mathematical formulation
there exist
formulation
a variety of
in
nature, or
adopted.
In
to operate
not amenable
the runway
the schedules.
necessary to force an aircraft to land
aircraft
are
in
to the
scheduling problem,
other operational constraints
controller may want to impose on
two
functions is a typical
world applications often impose constraints
either qualitative
mathematical
of constraints and
problem usually involves a degree
Linearization of higher order
In addition, real
are
attractive alternative for
We will discuss some of the most important ones.
objective function in an optimization
of idealization.
in TASIM has
that the ATM/C
For example, it may be
as soon as possible, or to force
direct succession.
Even if
the general
problem formulated in section 4.3 could be solved in polynomial time, it
would probably be impossible to incorporate the full repertoire of other
constraints which would be required
remains
in
control
of
to insure that the ATM/C controller
the automated
decision
support
system.
The
heuristic procedure can handle arbitrary forms of constraints as well as
arbitrary objective functions.
Seconds
at all
times prior
there exists a feasible solution.
to the
termination of
the heuristic
This is not the case with all optimal
-149-
algorithms.
first
In
feasible
solution
availability of a
dual algorithms, the
of them, the
an important subset
is
obtained
the
optimal
The
solution.
may be very important
feasible solution at all times
from an operational aspect. since within the overall system in which the
process
spaced
to
but
terminate abruptly.
occur
at
used
solutions.
When there
allowed
generate
scarce, strict
These events
stochastic
effectively
to
require the optimization
a number of events may
algorithm is embedded,
intervals.
in controlling
the
is plenty
better
are usually
time allowed
of time
property
for
When
the
can
be
reaching good
available the
solutions.
time limits may be
This
not evenly
algorithm is
time
available is
imposed on the algorithm
and we are
still assured of obtaining a good solution.
Third,
in
real
objective that should
problem at
are
hand.
all possible
world
applications,
be optimized.
Runway capacity,
The
problem
overwhelming majority
objective
from
that
of
the
a single
the case with the
costs. etc..
Exact algorithms often
of attributes that constitute a "good"
of real world
function that
seldom
aircraft delays, user
is compounded
near-optimal solutions exist.
is
This is clearly
candidates for optimization.
fail to capture the multiplicity
solution.
there
by
the
fact
that,
problems, a substantial
in
the
number of
Each may result in a value of the primary
is indistinguishable, efor practical purposes.
strictly
optimal
value.
Yet,
they
may
rate
(1) In the case of runway scheduling such an event may be a new aircraft
entering the system.
-150-
than the
higher
substantially
to
with respect
is designed to
The heuristic
objectives.
latter
the secondary
make intelligent trade-offs
among alternative solutions based on secondary objectives.
In
many
situations, including
algorithm
may
heuristic
procedures
adopting
be
the
only
may
mathematical
viable
easily
inferior solutions,
formulation,
the problem
alternative.
compensate
even when
can be
at hand,
The
for
the
the problem,
solved through
a heuristic
advantages of
possibility of
in its idealized
efficient polynomial
algorithms.
AL A
A
TL
~~~
Alf-----
The heuristic
adopted
[DEA 76].
by Dear
A
In
qr-o
atP'i 1
myn
algorithm implemented in
to generate
number of
TASIM uses the
feasible sequences
important
in a
improvements have
basic idea
systematic way,
been introduced,
however.
The scheduling logic has been extended to provide for multiple
runways;
time
varying scheduling
finally, the algorithm has been
constraints have
been incorporated;
adapted to the real-time environment in
which it has to operate.
The
heuristic
consists
of
three
parts
or
stages.
In
the
initialization stage, the problem parameters are set up according to the
"state"
of the
ATM/C system
at current time.
The second
stage is a
feasibility search which generates a feasible solution to be used as the
initial
solution
required
by
the third
stage.
The
latter
is the
-151optimization
heuristic
permutations
of the
none of the
which
generates
current best
new
sequence.
The
local permutations allowed by the
sequences
by
local
algorithm stops when
heuristic rule is better
than the current best.
4.8.1
The Initialization Stage
The initialization stage of the scheduling algorithm is depicted in
figure 4-1.
Upon being invoked the algorithm enters this stage.
the decision
aircraft set, Ad,
for rescheduling, A ,
rescheduling because
time
from
and the set of
are generated.
the current
time.
all aircraft ineligible
Aircraft may not
their current STAR
is within a
In addition,
its
position
and
may constrain a landing
its
surrounding
be eligible for
prespecified lead
the ATM/C
explicitly "freeze" the STAR for some aircraft.
generation algorithm
First,
controller may
Finally tpe flight plan
from being rescheduled if
traffic
pattern
do
not
allow
modifications to its flight plan to be generated.
The
scheduling
updated next.
constraints for
all
aircraft in
tne
system are
The updating procedure is related to the position of the
aircraft
along
assigned
to
its
previously
it. This
will
assigned flight
be discussed
plan
further in
and
the runway
connection with
automatic flight plan generation which is presented in the next chapter.
The time
next.
These
separations between all pairs
are
used
during the
of aircraft are calculated
feasibility
and
the optimization
-152-
Figure 4-1.
Runway Scheduling Heuristic, Initialization Stage.
-153stages.
Special
want to impose
separation requirements that the
on specific aircraft are taken
ATM/C controller may
into account during tnis
calculation.
The most common reason for
be
the
entry
of
a new
incorporated
in the
flight plans
for them, and
The
flight plan
nominal
invoking the scheduling algorithm would
aircraft
in
the system.
optimization process
aircraft are
New
by first
generating nominal
by inserting them in
the current schedule.
generation is based on
prespecified nominal
landing
and
routes.
Nominal flight plan generation will be discussed further in the
departing routes
as
well as
nominal speeds
along these
next- chapter.-
Landing aircraft
in the sequence.
be
at
the end
feasible.
is not
are initially inserted in
In most cases,
of
their nominal position
the nominal position for landings will
the sequence
and
the resulting
schedule
will be
It is possible, however, that the aircraft's nominal position
at
the
end
of the
current
schedule.
This
siLuation will
typically exist when travel times from various entry fixes to the runway
thresholds differ substantially.
schedule will be infeasible.
aircraft
is moved
In this
case, it is possible that the
Whenever the schedule is not feasible, the
backward in the
sequence until
a feasible schedule
results or until it reaches an infeasible position.
Of course, when two
or more runways are active, the
on
each active
runway are tried
nominal as well as subsequent positions
before moving
the aircraft backward.
-154-
all other aircraft is not changed by
Note that the relative sequence of
the insertion.
time required for landings
than the
to reach the runway
from an entry
An attempt is made to insert new departures in the sequence as far
fix.
forward as is allowed by their PTAR.
enough to
large
allow the departure
to this
assigned
sloc.
The
during
the
optimization stage.
active
all
possibilities
are
If a
gap between two landings is
to take off
without changing the
landing, the departure will initially
existing scheduled time of either
be
much shorter
which is typically
on taxi time,
runway is based
the
and their preferred time of arrival
difference between the current time
at the
because
differently
somewhat
treated
are
departures
New
assignment may
Again
tried
when two
before
the
of course
be changed
or more
runways are
departure
is moved
backward in the sequence.
Finally, once
schedule the constraint
that
have
generator.
aircraft have been
incorporated in the
list is updated to include
any new constraints
all the new
been imposed
by
or the
flight plan
obsolete constraints (e.g.
ones having
the ATM/C
During this process
controller
to do with aircraft which are not in the decision set Ad) are deleted.
4.8.2
The Feasibility Stage
In general,
the
algorithm
the schedule generated by
will not
be feasible.
the initialization stage of
The algorithm
will, therefore,
-155-
enter the feasibility stage.
procedure as the
same heuristic
the
generate the optimal
stage to
during the
in the optimization
The only
difference is that,
schedule.
optimization stage, the
that its value
are used
to compare schedules
feasibility stage uses
a specially
This
function is
Di infeaaibility.
called the index
defined function
defined so
one employed
objective function(s)
the normal
while
feasible schedule is generated by using
A
is zero when
positive if the schedule is infeasible.
the schedule is feasible, and
In essence, the method used to
to the standard method employed to
generate a feasible schedule is akin
generate a feasible solution to a linear program.I
A
variety of
forward
method
infeasibilities.
each constraint
can be- defined.
such functions
to
is
define the
here we
By infeasibility,
is violated.
Of
as
objective
The
most straight
the sum
mean the
all the
amount by which
course, the constraints
violated do not contribute anything to this sum.
of
that are not
As an example, suppose
that according to a schedule
STAR.
in
for
some
infeasible
aircraft i
and
the
and
> LFTAR.
in
some runway
contribution
of
n.
the
The schedule
scheduling
is obviously
constraint
for
aircraft i to the infeasibility index would be equal to STAR. -LFTAR.
.
(1) See for example [SIM 66].
in
in
-156-
the
point
of
value
and
is feasible
schedule
corresponding
infeasibility
the
If at any
infeasibility is least.
the index of
one for which
is the
current best schedule
feasibility stage therefore, the
During the
zero
tne
enters
the
to
goes
index
the. algorithm
optimization stage with that schedule as the initial solution.
The Optimization Stage
4.8.3
The
optimization stage
section 4.5.
The primary motivation
0-1 variables
of the formulation
easily constracted based
can be
algorithm constructed in
on the
is based
is that, given the
the optimal schedule
in section 4.3,
on a scheduling
values of the
assumption consistent
with the objectives.
In
the optimization
practice,
generation,
sequence
steps:
evaluation.
is performed in
generation,
schedule
The flowchart of figure
three sequential
and
the schedule
4-2 depicts the combined operation
of the feasibility and the optimization stages.
The method
for generating sequences of
is independent of
(current
best)
the number
schedule
of runways
is used
to
operations according to their STAR's,
on a
small
single runway.
subsequence are
At
any time, the
operations to be evaluated
that are
define
the
active.
base
The input
sequence
of
as if all aircraft were operating
positions of aircraft
permuted to produce
the new
sequence.
within a
We'will
-157-
YES
Figure 4-2.
Runway Scheduling Heuristic, Feasibility and
Optimization Stages (continues).
-158-
YES
YES
Figure 4-2.
Runway Scheduling Heuristic, Feasibility and
Optimization Stages (continued).
-159-
call this the
active subsequence.
The positions of
aircraft which are
not within the active subsequence are not affected.
The procedure begins by defining
the active subsequence to include
the last K aircraft in the base sequence, where K is an input parameter.
The next permutation within the
and checked
current active subsequence is generated
for feasibility with respect
to the sequencing constraints
relative to the overall sequence (constraints 3c and 3d of section 4.3).
If all the permutations within
evaluated,
aircraft
the subsequence
in
the
the current active subsequence have been
slides forward.
subsequence
becomes
The position
permanent
immediately preceding the-first aircraft
and
of the last
the
aircraft
of the subsequence is included
in its place.
Once
schedule
a
and
scheduling
new sequence
has thus
runway assignment
assumption.
This
can
been generated,
also be
is done by
the corresponding
determined based
on the
sequentially considering each
aircraft, starting with the first aircraft in the current permutation of
the active subsequence and ending with
sequence.
Note that
the last aircraft in the overall
only the schedule of aircraft
succeeding the ones
that have been resequenced could be affected.
If the
runway assignment of
not yet become permanent,
(1)
the aircraft under
the scheduled
consideration has
times of arrival at the runway
Runway assignments
usually will become
permanent
scheduled time of arrival at the runway is frozen.
before
the
-160-
are evaluated for all possible
runway assignments and the earliest STAR
is used.
may be infeasible due
A runway assignment
requirements within the runway, (constraints
In
addition,
reasons.
aircraft in
2a and 2b of section 4.3).
a runway
assignment
may be
infeasible
For example.
the runway
may be
too short
question, or the
to the sequencing
controller may have
for operational
for the
type of
explicitly requested
that aircraft of that type should not use a certain runway.
Given the runway
runway is determined
respect to all
In"
adA.Jdion,
iL
assignment, the scheduled time of
arrival at the
so that it satisfies the
spacing constraints with
aircraft for which STAR's have
already been determined.
to
has
at the runway EFTAR. .
in
WpLyiU
com
tle
earLiest
feasible
time
of
rival
STAR. is therefore given by:
in
STAR.
in
max {EFTAR.
in
,
t.
in
}
where,
tin
The maximization
{STARRWY(j)
3max
jiRWY(j)n
determining t. is performed
in
preceding i in the sequence under consideration.
over all aircraft
j
If at any point during
the schedule generation, any of the remaining constraints is found to be
violated, the process is stopped and a new sequence is generated.
The
evaluation
of
each
schedule
generated
captures
multi-objective nature
of the runway
scheduling problem.
alternative
functions are
"active" simultaneously.
time, the
objective
ATM/C controller can
decide what the
A
the
number of
At any
relative importance of
-161-
each
objective should
active
aircraft
delay
capacity
of
may
the
of
objective.
tertiary
the
be the
may
absolute number
secondary
of the
objective, and
may
be the
based
on the
position shifts
of
be
will presumably
The' decision
the average
maximization
objective,
primary
runway system
the
minimization
be
minimization of
Thus,
be.
current traffic situation in the terminal area.
is first done
is
used
to
objective.
is determined.
of all active objectives
The value
The secondary objective
based on the primary objective.
break
ties between
that
schedule
The comparison
minimize
the primary
tertiary objective is used, and so
If ties still exist, the
been- scanned.
on-until all -the-active-objectives-have
In the unlikely
event that ties still exist, the current best schedule is preserved.
Two schedules
are considered equivalent with
respect to a certain
objective if their values are within a certain percentage of each other.
This allows, for
of
thus
2.7 and
their
example, two schedules which result
2.8 minutes respectively,
performance with
deciding factor.
to be
respect to
The actual "margin
schedules are considered equivalent can
in average delays
considered equivalent and
other objectives
of equivalence" within
becomes the
which the
vary depending of the objective
in question.
The new
better.
schedule replaces the
current best if
it
is found
to be
When this happens, the sequence generation process is restarted
with the last
K aircraft comprising the active
subsequence once again.
-162-
terminates when the forward sliding
Accordingly, the optimization stage
is not in the
subsequence will include an aircraft which
of the active
decision aircraft set.
4.9
Implementation Status
the previous section has been
The heuristic algorithm described in
as well as
can handle both landings and takeoffs,
The software
provided by TASIM.
time environment
in the real
is currently operational
implemented and
the features which
Some of
multiple runway configurations.
were described in the previous section for the sake of completeness have
not
however, because
been~implemented,
software for
flight plan generation,
the controller
the existence of
they require
as well as
full specification of
These
scheduling software.
interaction with the runway
are:
1. Software
scheduling constraints applicable
determining the
software
This
aircraft.
with
conjunction
and
plans
flight
nominal
generating
for
Currently nominal
the
will
Traffic
be
Plan
flight plans are obtained
to each
fully in
implemented
Flight
for
Generator.
as input data
and are specific to each entry fix.
2. Software
to
constraints.
accept
and
The heuristic
variety of constraints.
process
controller
is designed
to allow
imposed
a wide
The full repertoire of constraints
-163controller, however, has
be available to the
which should
not been specified.
3.
however, separations should be
a fully operational system,
on
categories and
actual
the
of
classification
eliminate
In
of research and development.
adequate for the purposes
based
This is
as input data.
the separation minima are obtained
determined
Currently
the separation minima.
Software for determining
This
traffic.
prespecified
into
aircraft
will
runway scheduling algorithm
will allow the
to treat each aircraft as unique.
testing of the runway scheduling
There has been limited simulation
heuristic.
a'very practical method for obtaining
It appears to provide
an efficient
testing is needed, however, in
demonstrate
very high operational
runway schedule at
the
in
terms
Further
its performance and
order to establish
gains
efficiency
rates.
of
increased
runway
operational capacity, and reduced delays under given traffic conditions.
4.10
Coordination Among Terminal Area Control Sectors
Automatic
current
ATM/C
implementation
runway
scheduling represents
procedures
of
this
in
the
function
a drastic
terminal,
will
require
area.
departure from
In
particular,
restructuring
of the
-164coordination
procedures among
the terminal area
control sectors (i.e.
approach control, departure control, ground control,
In
the present
landing clearances,
system, the Tower
but does not
is responsible
have direct control
stream it receives from final
approach control, or
stream it receives from ground
control.
is responsible for
for sequencing
and Tower).
over the landing
over the departure
This means that Final Approach
sequencing landings, Ground
takeoffs, and the
for takeoff and
control is responsible
Tower is responsible
for interlacing
the two types of operations.
With
above
the
decisions
implementation of
runway scheduling,
are centralized.
Assuming
all three
for the
moment
of the
that the
current controller positions in the terminal area remain distinct, it is
clear
that substantial
this point
coordination will
we cannot provide an
be required
answer as to how
among them.
At
this should be done.
We will, however, briefly describe one possible scenario:
The
Tower
controller
decisions regarding the
the software, he
current
has
primary responsibility
use of the runway system.
generates the runway schedule.
responsibilities,
i.e.
assuring
for
all major
With the support of
He
the safe
also maintains his
operation
of the
runway system.
(1) Here ground control is the terminal area sectoZ responsible for
controlling traffic on the surface of the airport and n=t about the
ground control system in general.
-165-
interacts with the runway scheduling
The Tower controller normally
establishing the
of
the purpose
for
software
general
strategy.
He
decides, for example, which objective function should be optimized, what
MPS value is appropriate for the particular time of
constraints (e.g.
tactical
impose
however,
He may,
day, etc.
specifying
tne position
aircraft, requiring that two aircraft exchange
and/or the STAR for some
the software to schedule some landing
position in the sequence, forcing
at the earliest possible time, etc.) if the need arises.
The
controller also
Tower
makes
informs the scheduling function of unusual
configuration to be used and
situations that have to be taken into account.
missed approach for rescheduling,
that other controllers
the runway
decisions regarding
he enters a
For example,
Finally, he reviews constraints
etc.
may have have imposed and
negotiates changes if
required.
The current
schedule for departing
and
controller
Ground
to
the
aircraft is made
Departure
known to the
controller.
The
Ground
controller is responsible for the management of the departure queue.
He
has to insure that the order in which aircraft are queued at each runway
holding
area
is the
one required
schedule is available, the ground
gate
hold
based
on the
length
decisions can
now be made
hold interval
will depend on
by the
schedule.
Since
a precise
controller does not have to implement
of
the departure
on an aircraft-by-aircraft
the scheduled delay and
taxi time for each individuai aircraft.
queue.
basis.
Gate hold
The gate
on the estimated
-166The Final Approach controller obtains scheduling information on all
landing
aircraft.
He is responsible for
delivering landings
Outer Marker at the time implied by the schedule.
schedule
by
imposing
schedule cannot be
constraints
whenever,
safely implemented.
in
at the
He interacts with the
his
judgement,
He is, therefore,
the
the one most
likely to use the tactical constraints which were mentioned above in the
discussion of the responsibilities of the Tower controller.
Finally, the Entry cnntroller is responsible for landing traffic as
it first enters
on
the
holding
schedule
stacks
the terminal area.
for landing
(when
they are
He, also,
aircraft
has complete information
and manages
needed)
in
controller manages gate hold operations.
the same
the
way
operation of
the Ground
It is important to note again
that runway scheduling allows holding stack
management to be done on an
aircraft-by-aircraft basis.
4.11
Computer Aided Vectoring for Approach Spacing
It is impossible to
excpect that a
fully
automated ATM/C
system
can be introduced at once.
For this reason, it is important to consider
the
when
evolutionary
period
automation
functions
are
gradually
in order to determine the interim needs of air traffic controllers.
In order to
exemplify
how TASIM
can be useful in addressing this
type of problem. a special purpose display was designed and implemented.
-167-
The
display is designed
final
spacing
runway without
provides the
The display
visual method to time the final
controller with a
controller in
aircraft at the
flight planning.
of automated
support
assist the
delivery of landing
achieving accurate
the
to
turn onto the runway
centerline.
of the final
The geometry
of
part
final vectoring
the
vectoring
display
figure 4-3)
(shaded in
area
approach area, shown in
3-8).
The final
a point
called the
(also figure
starts at
figure 4-3, is
vector marker (VM) which is situated about 2 miles from the outer marker
(OM).
The
glideslope.
to
OM
is
the
point
where
It is assumed that the
the runway
aircraft
would acquire the
final vector will be less than +200
that the
centerline so
the
pilot/autopilot may acquire and
commence tracking the centerline with less than a 200 change in heading.
The
displacement of
the VM is
designed to allow
stabilization on the
runway centerline by the time the OM is reached, even if the aircraft is
vectored along the edge of the
final vectoring area and thus intercepts
the centerline at the VM.
Horizontally
representing
The
number
across
arrivals from
inside
each
the
top
of
the left or
box indicates
the
display,
the right side
the
position
there
are boxes
of the runway.
in
the landing
sequence currently assigned to the aircraft by the runway schedule.
boxes move towards
the landing
the runway centerline extension at
speed of the
aircraft they represent.
The
a speed equal to
-168-
19
f2j_
49
VECTOR
MARKER
OUTER
MARKER
Figure 4-3.
Final Vectoring Display.
-169-
a box reaches the runway centerline extension,
When
1. it expands to a size
error in the
which indicates the maximum tolerable
delivery to the runway.
This
value has been
specified, and will generally depend on the accuracy of the
surveillance system.
2. it grows
a "wand" to
the left or right
(depending on the
position of the aircraft it represents). and
3. it starts moving along the centerline extension towards the
OM.
The wand
time to
200
is positioned so that
call the final vector
angle.
when it touches the
target, it is
to intercept the runway
centerline at a
If this procedure is performed properly,
the target will
intercept the box on the centerline at some point before the VM.
procedure has
This
only.
It can
the timing
provide the
speed.
currently been
mechanized for
be easily extended, however, to
of the turn to
Appendix
help the controller with
the base leg as well.
timing for calling
B examines
Finally, it can also
the deceleration to
how the
the final turn
positioning of
the final approach
the wand
can be
determined as a function of the aircraft current and landing speeds.
....
es
-,g,,.
.'-'
-170-
r 1 es--.-,
n.
. , ,
,,..
. s ...-.g'J;
r--
-171CHAPTER 5
A METHODOLOGY FOR TRAFFIC FLIGHT PLAN GENERATION
5.1
Introduction
operations provides the potential for
Optimal scheduling of runway
substantial improvements
the runway system.
in the efficiency of
potential can only be realized if the schedule is achieved.
that
landing
threshold
at the
delivered
must be
aircraft
This
This means
of their
assigned runway on time.1
To
precise
accomplish
conflict-free
plans
flight
the
at
delivery
runway,
generated for
are
all
4-dimensional,
landing aircraft.
These are based on the aircraft's current position. as determined by the
surveillance and tracking system, and the aircraft's STAR, as determined
by the runway schedule.
Fl-ight
by the
planning, performed
Traffic Flight
Plan Generator
(TFPG), establishes with certainty that the runway schedule is feasible.
there
i.e.
exists
a safe
(conflict-free)
set of
paths
which will
satisfy the schedule.
Flight
precise
planning
delivery
is
only
of landings
the
at
first
step
the runway.#
towards accomplishing
Once flight
plans are
(1) Obviously departures should also be on the runway ready for takeoff
at their scheduled time. This, however, does not involve flight
planning.
-172-
specified, there remains the problem of guiding the aircraft along their
through
is accomplisned
guidance
Aircraft
path.
assigned
the
interaction of the Command Generator, which determines a set of commands
for
suitable
and
the
and
Detection
Command
in a timely
these commands
and dispatches
Conformance
the
capabilities,
navigation
aircraft's
which activates
Activator,
fashion,
each
Conformance
Resolution
functions, which insure that the flight plans are indeed being followed.
Due
to
the close
guidance,
the
relationship
methodology
between flight
presented
here
planning
will,
on
and aircraft
many occasions,
encompass the guidance functions as well.
ConsisCte
wit
Ute
bi
design
philoSUphy
of
the auLomatied
terminal area ATM/C system, it is necessary to maintain the master-slave
relationship between the controller and the TFPG software.
that
flight plans
ATM/C controller.
the flight plans
or
legs.
The
should be
easily understood
This implies
and visualized
by the
Accordingly, the horizontal profile (ground track) of
will be composed of a small
third and
fourth dimension of
number of linear segments
the flight
plan will be
provided by specification of altitude and speed change points along some
of the legs.
Advanced navigation and flight control systems onboard the aircraft
will
enhance
however, be
the
operation of
a prerequisite for
the
ATM/C system.
using it.
profiles are also consistent with this
Piecewise
They
should not,
linear horizontal
design goal for the TFPG and the
automated terminal ATM/C system in general.
-173-
schedule.
changing runway
As
a
ground tracks are selected
To achieve this,
from a prespecified path structure
which allows a number of alternative
generated for an aircraft at any
paths to be
considerable
result,
until
exist
shortening
adapt to a dynamically
to generate flight plans that
The TFPG has
flexibility in
the
late phases
point along its approach.
path
of
stretching
and path
approach,
when the
the
aircraft's scheduled time of arrival at the runway becomes fixed.
The changing runway schedule has important implications with regard
to aircraft guidance as well.
any
when its
aircraft,
strictly
enforce
be
STAR will
changing, there
the -4-dimensional- flight
function is
Detection
the early part of the approach of
During
plan.
is no
The
need to
Conformance
identify non-conforming aircraft.
still used to
However, instead of forcing the aircraft back into conformance, the TFPG
uses estimates of their current position, ground track, and ground speed
to generate
new flight plans.
In
other words, in the
the runway schedule had
the approach we can treat non-conformance
as if
changed.
of the flight
Thus,
the inherent flexibility
early phases of
path selection
process is used to correct navigation errors, errors in the estimates of
the prevailing winds,
and finally, as a means
of avoiding active speed
control on most occasions.
Changes in the runway schedule
usual events
plan.
In
generated
that will result
some cases, however,
for some
and conformance alerts are tne most
in modification of
a revised flight
aircraft in order
to resolve
an aircraft's flight
plan will have
to be
conflicts which have
-174-
the ATM/C
controller may impose
issuing a
direct command to
Finally,
plans of other traffic.
resulted from modifications of flight
plan, either by
changes in the flight
the pilot, or
by specifying a
new flight
plan for the aircraft to the TFPG function.
The implication of the above
scenario is that, flight planning for
the terminal area ATM/C system is accomplished by repeatedly solving the
2.3.3).
point in time when new flight
At each
(i.e.
runway
threshold) for
Q%
~
precisely
spciie
Oyji-h
.OZA.
...
~
Q"S
each landing
~
~
in
J.L&
-r
dienios
U.UL
~~rr4
~
.LAd"'.
plans are required, the
and the
position)
the current
origin
ri.
chapter 2 (section
plan generation problem stated in
same basic flight
destination
aircraft in
*A.
lgtpanigts
A...L&"J-L
Th
JLLI=
the terminal
r.L
""J.
"6L
the
(i.e.
area is
%.CLO'%
st
.L0
6_0
find a set of 4-dimensional, conflict-free paths which satisfy the above
boundary conditions.
The link between consecutive invocations of the traffic flight plan
generation
function
positions (i.e.
point in time.
pointed out.
options
them
solution
itself.
Namely,
the aircraft
TFPG) are a direct consequence of the
the output) which
were generated
at some earlier
Even though this relationship is obvious,
it has to be
It emphasizes that
flight plans must leave
open as many
in order to
allow modifications. if
the need for
as possible
arises
the
the input to the
plans (i.e.
flight
is
at some
future
time.
This
flexibility
needed in the initial phases of the approach.
is particularly
-1755.2
Flight Path Structures
A
typical layout
figure 5-1.
The
of the Terminal
Control Area (TCA)
TCA extends 30 to 60 nautical
which is situated in the center of the figure.
the periphery of
Landing
miles from the airport.
A number of waypoints at
the TCA have been designated as
traffic is directed'by
is shown in
entry and exit fixes.
enroute controllers
towards the entry
fixes* (shown as upward pointing triangles) along jet routes or airways.
Departing traffic is directed towards the exit fixes, which are shown as
downward
pointing
triangles.
The
based on the destination airport
exit
fix assignment
of the departure.
is determined
Landings are handed
off to the ATM/C controller 5 to 10 nautical miles prior to reaching the
entry
fix.
Departures
are handed
nautical miles prior to reaching
from the
entry fixes to
solid lines.
off
to
the exit fix.
the runway are
The dashed lines represent
enroute sectors
5
to 10
Nominal approach routes
shown in figure
5-1 in
nominal departure routes from
the runway towards the exit fixes.
The nominal approach route from entry
in greater
detail in figure 5-2.
fix A to the runway is shown
It consists of 5
legs.
1. The Entry leg, (OA in figure 5-2)
2. The Initial Approach leg, (AB in figure 5-2)
3. The Downwind leg. (BC in figure 5-2)
4. The Base leg. (CD in figure 5-2)
5. The Final Approach leg, (DE in figure 5-2)
linear segments or
-176-
TCA boundary
)
>
/
V
ACM
V
7IV
.z
x
-~--
'I
I
V
Figure 5-1.
-
V
Typical Terminal Area Layout.
.~'-. -
A
-177-
20NM
* FAS = Final Approach Speed
15NM
D
C IAL T =2 ,000 f t I
-
IAS=160 kts
10NMALT=2
IAS=F,)
rE
5NM
ft
M
B{ALT=4,000
ft~
IAS=180 kts
P
TCA boundary
ALT=15,000 ft
.IAS=220 kts
0
Figure 5-2.
ft
IAS= 250 kts
{ALT=15,000
Typical Nominal Approach Flight Plan.
-178-
Note
the leg
that
from point
is also
Furthermore, we have
leg.
exist until the
will continue to
an outer marker
marker, (OM),
the outer
part of the final approach
considered to be
assumed that
E to
ILS is
decommissioned.
leg is associated with a
Each
area
airspace, and
with the
same general
location of the entry and downwind
traffic
in
them,
downwind,
base,
specific
to each
base
regions
depends
on
and final
direction of
traffic.
The
regions, as well as the direction of
the location
of
the
regions, on
approach
This
runway direction.
are merging
of the terminal
particular region
points for
fix.
the other
means that
traffic from
entry
The
hand, are
the downwind and
two or
more entry
fixes.
From nominal approach routes, we can construct nominal flight plans
by
specifying
flight plans
altitude and
are based on
speed
profiles along
typical values for the
rates at the preferred aircraft configuration (e.g.
from
"idle thrust"
descents at various
they may vary depending on the
altitudes and
leg.
Nominal
airspeed and descent
airspeeds resulting
flap settings).
aircraft type.
indicated airspeeds, (IAS),
each
Consequently,
Figure 5-2 shows typical
at various points
along the
nominal approach route for jet aircraft.
Nominal flight plans
can be thought of as
the ideal 4-dimensional
flight profile from the entry fix to the runway for each type of landing
aircraft.
Thus, they determine the preferred time of arrival at various
-179intermediate waypoints as well as PTAR, the preferred time of arrival at
aircraft.
for each landing
the runway
we have already
The latter, as
for the runway
the sequencing constraints
noted, is used to implement
scheduling function.
each leg of the nominal
The TFPG associates
phase
The
approach.
af the
P
5-2 )
figure
in
the flight planning options that are
may
and fly along a downwind leg
Similarly, a
the' figure.
soon as the
As
aircraft.
significant
transition
downwina
at
at
phase at
points P
or
which is parallel to the one shown in
choice of
base legs
is available
aircraft transitions to the
its flexibility
part of
to the
the runway centerline, (e.g.
different distances from
P2),
is when its
An aircraft which is in the initial approach phase (e.g.
available.
point
the aircraft
phase in which
flight plan is revised, determines
route with a distinct
downwind leg, a
since only
is lost
to that
the base leg
selection remains available.
The
performance of
have
which
runway.
the ground
geometry of
the TFPG, particularly with regard to
to achieve
Many
important effects
tracks has
of the
a
precise and
characteristics
landing aircraft,
schedule at
tight
we will
in the
discuss,
the landing
however, are
equally as important in flight planning for departing traffic.
Eirst, in order to avoid the accumulation of errors as the aircraft
proceeds
along its
aircraft
at
flight plan,
intermediate
small errors
waypoints
must
be
in the
delivery of the
readily
absorbed
by
-180modifications during
approach
initial
the ground track.
subsequent legs of
and
downwind
the
this
phases,
capability
of
approach, it will
the
determine the
will
During the later
substantially limit the need for strict speed control.
phases
During the
accuracy
with which
landings can be delivered to the runway.
Secod,
a multiplicity
aircraft to the runway at the
will
insure with
of
ground tracks
capable
of delivering
scheduled time should be available.
high probability that
a set
This
of conflict-free flight
plans exist.
Third, scheduling
along
the flight
flexibility should be maintained
plan to
the highest
at every point
degree possible.
The interval
between EFTAR and LFTAR will generally decrease as the aircraft proceeds
along
the various
phases of the
approach.
difficulties since rescheduling during the
anyway.
Some
flexibility needs to be
maintained, however, even during
to allow small schedule
by failure of preceding operations
changes brought about
In
not present any
late phases is not desirable
It will be necessary in order
the base phase.
schedule.
This does
many respects, this
to meet their
characteristic is equivalent
to the
capability to absorb delivery errors at intermediate waypoints.
In
order
to
obtain
these
structures which provide a number of
the
approach.
accomplish this.
There
are
many
characteristics,
we
specify
path
alternative legs for each phase of
alternative
structures
which
will
To be concrete, one such possibility will be described
-181-
(figure 5-3).
solid line depicts the nominal
The
approach route which
is identical to the one shown previously in figure 5-2.
approach requires somewhat separate
each phase of the
Even though
treatment, we can distinguish two major stages during which the goals in
The first stage, which we call
generating flight plans differ greatly.
adaptive.
entry, initial approach
and includes the
incoming flight
intormation on an
the ATM/C system first obtains
starts when
and downwind
The second, or precision. stage includes the base and the final
phases.
Each
phases.
approach
from the
aircraft transitions
adaptive to the
precision stage when its scheduled time of arrival at the runway becomes
fixed.
The dashed lines in figure
5-3 represent the alternative legs that
Thus, for the initial approach
are available during the adaptive stage.
phase, the aircraft can be routed along leg AB or along any leg parallel
to AB.
for
For example, the ground track AA B
this
of
phase
the
Legs
approach.
is an acceptable alternative
to
parallel
BC
provide
alternatives for the downwind phase.
stage, the aircraft's
During this
flight
goal,
planning
therefore,
STAR is changing.
is to
maintain
the
The primary
time interval
[EFTARLFTAR] as large as possible so that runway scheduling flexibility
is maximized.
there
will be
Due
to changes
cases where one
during this stage.
in the
sequence of
aircraft will need
Since all aircraft will be
runway operations,
to overtake another
descending while in the
20NM
-182-
-
15NM
-
D
C
10NM -/I\
E
B
K
L
Figure 5-3.
xl
_
NAN_
I
Example of an Approach Path Structure.
-183and
initial
altitude
phase, it may not
the downwind
between overtaking
separation
therefore, need
to be on two
the initial
aircraft will,
In order
to insure that
between overtaking traffic, adjacent legs
there will be no interference
in
Such
traffic.
different legs.
to maintain
be possible
3-5 nauricai miles
be spaced
phase will
and downwind
apart.
The selection in the base phase is made from a number ot legs which
are
perpendicular
to the
approach phase, the legs intercept
small angle (typically 200).
In
extension.
runway centerline
the final
the runway centerline extension at a
During the base and final approach phases,
the aircraft's STAR is fixed- and flight planning is primarily concerned
precise delivery
with
cannot
be achieved
spaced
3
or 5
of the aircraft
if ground
nautical
approach legs do not occur
plan
generator
will
to the
restricted to
tracks are
miles.
Thus,
the
turns to
at prespecified points.
determine
the time
and
aircraft should transition from the downwind
that to the final approach phase.
This precision
runway.
legs that are
base
and final
Instead, the flight
position
at
which the
to tne base phase and from
One such selection is shown in figure
5-3.
In
order
to
achieve
the
operation of the automated ATM/C
precision
necessary
for
the
proper
system, the timing of the deceleration
to the final approach speed will be controlled during the final approach
phase.
The final
approach will, therefore,
be the
only phase during
-184-
which active speed control will be exercised.
At all other times, speed
control will be enforced only if no other alternative is available.
The structures used in this particular geometry were chosen because
they
are widely
ATM/C system.
used for manual
spacing in the
The structure for
the initial approach phase is commonly
called a "harp",
while the structures for the
approach
collectively
phases
describing
the
present terminal area
form
a
downwind, base and final
"trombone".
operational characteristics
of
Analytical models
these and
other ATM/C
structures have been developed by Simpson. [SIM 64].
3
Ground Track Selection and Speed Control
methodology for ground track selection
We will present the general
and speed control using the selection
example.
of the initial approach leg as an
The method of selection for subsequent approach phases follows
the same principles.
Let
approach.
point
us consider
an aircraft
Referring back
0 flying
have
to figure 5-3,
towards the
specific initial approach
specified the
entry fix
leg.
the entry
phase of the
the aircraft may
A.
We
first have
This selection cannot be
legs and the
phases of the approach.
which is in
speed changes for
be at some
to select a
made until we
all the subsequent
The selection problem can, therefore, be viewed
from a different perspective.
Namely, we
desirable flight plan for the
downwind, base and final approach phases.
can first determine what is a
-185and then select
an initial approach leg based on
that flight plan.
We
will call such flight plans desired flight plans.
In most cases, desired flight plans will be very similar to nominal
flight
plans,
profiles etc.
since
flight
pertain
to
preferred
airspeeds, altitude
However, we prefer to use distinct terms because:
nominal flight plans
desired
both
encompass all the phases of
plans only
include
a specific
first,
the approach, whereas
subset of
phases, and
second, nominal flight plans are preferred from the point of view of the
aircraft pilot alone, while desired flight plans take into consideration
scheduling and flight planning goals as well.
A concrete
example will help illustrate
in the determination of the desired
additional
considerations which
the major issues involved
flight plan for any aircraft.
pertain to
Some
specific landing sequences
will be discussed in section 5.6.
For the final approach and the base phases, the desired flight plan
may be the
same as the nominal.
This
distance of 12 nautical miles from
means that the base leg
is at a
the runway threshold (d nom in figure
5-3), the turn to final approach is such that the intercept point E is 8
nautical miles
from the runway
final approach
airspeed occurs immediately following the acquisition of
threshold, and the
the runway centerline extension.
The
that
approximately
the
control
points
respective selection range.
are
deceleration to the
rationale for these selections is
centered
within
their
It is assumed, therefore, that the base leg
-186can be
between 6 and
at distances ranging
15 nautical miles
from the
runway threshold, and
that the turn to final approach
can be such that
the intercept point E
can range from very near the OM
to very near the
point
of
intersection
of
base
the
and
leg
runway centerline
the
extension.
Similar considerations apply to the selection of the desired flight
plan
for the
downwind phase.
desired flight plan
occupies in
"CentPringr"
the
the runway
the
.
considered also.
point to turn
in this case)
of the aircraft in the current runway
other hand, the aircraft has been
If, on the
to be
its nominal position. POS nom
the same as
In
however, the current position the
sequence has
nnie-
rntrol
would be adequate if the position
schedule is
nominal speed profile.
can be identical to the
determining the desired downwind leg,
aircraft
profile for the
example, the speed
For
(see chapter 4).
moved forward or backward
from its nominal position, its potential for being moved again by future
rescheduling is no longer the same
assume an MPS
value of 5, and that the
backward 3 positions.
at
most
in the two directions.
aircraft has already been moved
It can only be further
two positions.
At
the
same
For example,
time,
rescheduled backwards by
it is possible
to be
rescheduled forward by as many as 8. This implies that the selection of
the
downwind leg
figure 5-3,
has to
be biased
the downwind legs
accordingly.
which are at greater
runway centerline would be preferable to leg BC.
In
the situation of
distances from the
-187Once
the
desired flight
plan
for the
approach phases has been selected, we
initial approach,
*that
the
downwind
the point where
aircraft reaches
leg at
the
downwind, base
can determine, for any leg of the
a speed command
has to be
intersection with
the appropriate
and final
time.
The
the
given so
already chosen
initial approach
leg for
which the timing of the speed reduction most closely matches the nominal
speed profile will be selected.
speed
reduction,
deceleration can
nominal
In
speeds
be assumed.
the computation of the time for the
for
both
For example,
before
entry airspeed of about 220
in
180
vicinity
Similarly,
of
during the
knots
during
downwind phase
after
the
typically jet aircraft will
decelerate from their
the
and
the
knots to an airspeed
initial
aircraft will
approach phase.
lower their flaps
which will cause approximately a 20 knot decrease in airspeed.
The flight planning logic applies limits to the timing of the speed
change in
each phase of
knowledge
on
when (or
the approach.
at what
approach phase in question.
5-3 for example,
point) speed
desired
flight
associated
speed
conflict check.
limits reflect empirical
changes occur
along the
During the initial approach phase of figure
we may specify that the speed
plan should occur within 10 miles from
the deceleration,
These
the entry fix.
as determined above, occurs
plan together
with
change is accepted
reduction in the flight
If the timing of
within those limits,
the initi4l
approach leg
pending altitude
the
and the
assignment and
-188-
In
allowable limits.
to
aircraft and
the speed reduction
Having
initial approach
appropriate
leg.
the chosen leg
flight plan,
the desired
latest) allowable.
the
not
will
fall
within the
This simply means that the desired flight plan cannot
as is. Since
be implemented
respect
speed reduction
the
cases,
some
leg and
has been the
however, it is assigned
is set to
occur at the
thus selected the leg and
phase only,
the TFPG
the corresponding
The selection is again based
"best" with
to the
earliest (or
speed profile for
will proceed
speed profile
to select an
for the downwind
on the final approach and base phase
portions of the desired flight plan.
This iterative procedure continues
until the complete flight plan has been generated.
5.4
Altitude Selection
Aircraft
required
altitudes
separations
separation cannot
be
cleared
to
will
among
be
aircraft
be provided.
the
lowest
actively controlled
whenever
Within those
possible
to
adequate
achieve the
longitudinal
limitations landings will
altitude
consistent
be taken
into
with their
position and distance from the runway.
Obstructions
on
the ground
can
consideration by
imposing lowest safe altitudes along each leg in the approach structure.
Maximum altitudes
along the ground
track will also
be imposed.
These
have to be consistent with typical descent rates for each aircraft type.
The upper altitude limits will,
Unlike
lower altitude
therefore, depend on the aircraft type.
limits, upper limits
are not
associated with a
-189specific location
to the
within the terminal area.
time left until
limits applicable to
substantial change
touchdown.
Instead, they are related
This means that
any aircraft will have to
in its STAR.
the upper altitude
be redetermined after a
The reverse may also
occur:
on rare
occasions, the current altitude of an aircraft may affect the aircraft's
EFTAR.
5.5
Conflict Identification and Resolution
As stated in
to
section 5.1, in order for
operate properly,
mean
that
flight plans must
maintained at all
times.
traffic, but
airspace.
be conflict-free.
if-all- aircraft-. -adhere to- their -- flight -plans
specified conformance limits, the
other
the automated ATM/C system
In terminal areas,
within the
required airborne separations will be
An aircraft may be in
also with
By this we
conflict not only with
ground obstructions
and with restricted
airspace restrictions will usually result
from weather conditions. (e.g.
thunderstorms concentrated at particular
regions of the terminal area).
Given
the
identification of
conflicts
with
straight forward.
separation
requirements
conflicts is a straight
ground
In
obstructions
and a
set
of
flight plans.
forward task.
Resolution of
and restricted
fact, as we pointed out
airspace
in the previous section,
altitude
restrictions
process.
The same can be done with airspace restrictions.
are
inherent
to
the
is also
flight
plan
generation
-190-
Conflicts
flight
between
plans
of
determine which
two aircraft
either
(or both)
is the best
aircraft have to
can be
resolved by
aircraft
involved.
In
choice, alternative flight
be generated and evaluated.
These
changing the
order to
plans for both
may be in conflict
with the flight plan of a third aircraft, and thus a new decision of the
same
type
has
to
be
made.
This
approach
will
require excessive
computational effort and has to be rejected on that basis.
are
A much simpler conflict resolution
logic is adopted.
classified into
invalid,
Invalid
three categories:
flight plans
are those which
Flight plans
temporary,
do not conform
to the schedule.
This caLtegUry incLudes aircraft LhaL have been rescheduled,
of
conformance,
assigned flight
as
well as
plans.
aircraft
plans
are
which have
Temporary flight plans
the runway schedule but have not
flight
that
conflict-free
and final.
aircraft out
not
yet been
are in conformance with
yet been checked for conflicts.
and will
not
change
Final
during current
invocation of the flight plan generation function.
All three categories of flight plans have fixed portions.
Usually,
those will be parts of the flight plan which have been executed, or will
be executed within
a short time interval (say.
time the flight plan generation
of flight
controller.
plans may be
30-60 seconds) from the
function was invoked.
fixed due to
Additional parts
constraints imposed by
the ATM/C
When the flight plan generation function is invoked, invalid flight
plans are
the
identified.
function
temporary.
to be
At least
one flight plan should
invoked.
All valid
flight
be invalid for
plans are
then made
Finally fixed portions of all flight plans are identified.
Conflict checking is performed as follows:
1. Select an aircraft whose flight
If the
flight
plan is
plan is temporary or invalid.
invalid,
generate a
valid
one and
designate it as temporary.
2. Identify
conflicts with
final flight
plans as
well as with
fixed portions of invalid or temporary flight plans.
3.
If no conflicts exist, make the flight plan final.
generate a revised
temporary flight plan and go
Otherwise,
back to step
2.
4. If no more aircraft to consider stop. Otherwise go to step 1.
In practice, the flight plan generation and the conflict resolution
are performed simultaneously, i.e.
for
each
phase
alternatives
conflicts
of
are
approach, it is
generated if
exist, then
When complete,
the
as each the flight plan is generated
such conflicts
flight planning
the flight plan
checked
for
are identified.
continues with
becomes final.
conflicts and
This
If no
the next phase.
method guarantees
that conflicts will be identified at the earliest possible time.
-192This approach to conflict resolution has obvious drawbacks compared
with the more
general approach summarized earlier in
is possible,
for
conflict-free
flight
according
which aircraft
to
example,
to
be
unable
plans because
are
of
to
a bad
determine
choice in
selected in
therefore, very important to determine
this section.
step
a
set
It
of
the sequence
1 above.
It is,
a good criterion for making this
selection.
We have chosen
to select aircraft in an
scheduled time of arrival at the runway.
for aircraft which are close to
see,
lanin
ligt
time approaches.
This implies that flight plans
landing become final first.
fexbiltydereases
as th
Therefore, given two aircraft
land in direct succession, revising the
in most
increasing order of their
cases, be preferable
schele
As we have
lr
andingL~
which are scheduled to
flight plan of the second will,
to revising that
of the first.
This is
exactly what the conflict resolution logic will do.
5.6
Separations during the Base and Final Approach Phases
When the scheduled time of arrival
at the runway becomes fixed for
some aircraft, it transitions from the adaptive to
of the approach.
This will typically happen as
the precision stage
the aircraft is flying
along the assigned downwind leg.
Figure
5-4 shows
depicts a typical
an aircraft
on the
downwind leg
flight plan from there to the
(point P) and
outer marker (OM).
At
-193-
P1
P3
OM
Figure 5-4.
Typical Flight Plan for the Base and
Final Approach Phases.
-194-
points P
and P2 the aircraft
intercept
heading
runway
At the
allow easy
To
respectively.
angle 0
centerline extension,
degrees.
turns to the base leg
is typically
interception
of the
between 20
and 30
aircraft tracks
P3 , the
intercept point
and to the runway
the runway
and subsequently towards the runway
centerline extension towards the OM
threshold.
Aircraft have to
arrive at the OM not only
at a specific time but
also at a specific altitude and speed which will allow them to intercept
We will call
the glideslope.
this altitude the
5 nautical
is approximately
Typically,
the OM
threshold.
This means that the OM
DM crossing altitude.
miles from
the runway
crossing altitude is 1800 feet above
ground level.
If
aircraft is not at
the
altitude
command
will
be
the proper
given (at
aircraft will be cleared to decelerate
at some point
P .
s
some
altitude, an
OM crossing
point
P ). Finally, the
down to its final approach speed
P and/or P may occur
s
a
either before or after point
P2, and in any order.
This geometry is based on the
work of Durocher EDUR 77] who snowed
that it can provide excellent accuracy in delivering aircraft to the OM.
His
tests
simulator.
were
conducted
by
simulating
approaches
in
a
cockpit
Radar tracking errors as well as wind were also simulated.
-195this
In
section we
aircraft at the time of closest
to determine the separation between two
as a
approach,
the turn,
of
timing
required
he
runway.
planning
an li
adverse ects
region without any
Second, by
parameters
guidelines which can
on
studying
the
in the
be maintained
separations can
precision
be used to generate good
it
ni The delivery at
various flight
separations,
longitudinal
First,
final approach
of the
the effects
the
the minimum
is twofold.
Its purpose
the OM.
parameters (e.g.
e, etc.), and
intercept angle
the
separations at
that safe
shows
flight planning
function of the
model can be used
This
the final approach region.
traffic patterns in
that describes
mathematical model
present a
we
develop
traffic patterns in this
area.
Good traffic patterns cannot be established once the aircraft is on
the downwind or the base leg.
remaining
in
the timing
At that point, all the flexibility still
of the
needed to correct for navigation
the
desired traffic
during the entry
section,
desisred
turns and
Planning for
final approach area
has to start
patterns in the
and the initial approach phase.
plans which
are
deceleration is
and surveillance errors.
therefore, is intended to provide
flight
the final
used for
The analysis in tnis
guidelines for determining
flight planning
during the
adaptive stage of the approach (see section 5.3).
(1) The model is similar to the one first used by Dunlay for estimating
the expected number of conflicts at the intersection of airways.
[DUN 75]
-196-
Consider
the situation
We will
approach phase.
final
aircraft are
Two
5-5.
of figure
in the
that aircraft
assume throughout
1 is
scheduled to land before aircraft 2, and that both aircraft have already
decelerated to their respective final approach speeds, v
Aircraft
2
is flying
along
is going to intercept
Aircraft 1
and the intercept point P. is at a
the
centerline
runway
the runway centerline at
distance d. from the OM.
separation, S.
determine the minimum
and v 2 '
,
between the two
extension.
an angle 8,
We seek to
aircraft, as a
function of:
and v 2 , their final approach speeds,
1. V
2. S
,
the required minimum separation at the OM,
3. d., the intercept distance, and
4.
6 the intercept angle.
before aircraft 1, the minimum horizontal
If aircraft 2 reaches P.
3.
We will, therefore, analyze the case where aircraft
reach the OM first.
first.
1 reaches P.
3.
To simplify the formulas, we
the origin of the coordinate system and
aircraft
1
P..
is at
inconsequential.
aircraft 1 has to
be zero since, by assumption,
separation will always
We
Clearly,
the
will, therefore,
to be
will define P.
3.
we will assume that at time t=O
direction
assume
of
the
that both
flight
is
aircraft are
flying away from the OM.
The coordinates.
time t, are given by:
(x ,y ) and (x2 y)9,
of the two
aircraft at any
Y
AC #1
(x1 ,7
1
\
P.
AC #2
V.
x
Figure 5-5.
Horizontal Separations during the Final Approach Phase.
-198-
x
=
vItcos6
and
1 =
y
1 is at P..
When aircraft
1 is at the
when
OM, the separation
S. is, therefore given by:
between the two aircraft is S .
0
Si
0
two aircraft at time t=0, i.e.
where, S. is the separation of the
aircraft
=
5-1
2
i
2
1
= S
5-2
v2 )
(v-
-
V.
This
1 reaches P . first.
By assumption. aircraft
means that this
discussion is valid for:
d.
S(t),
.
Sv
o 1
(v1 - v
2
<
two aircraft
between the
the separation
as a
function of
time, satisfies:
(X1
S(t)2
Substituting equations
-
x2 2 + (yl - y2 )2
5-1 for the aircraft
coordinates we obtain
after some manipulation:
2 E V12 + V 2 2
S 2 (t) =
+
2tS.[
v2
-
V1
-
2v 1 v 2 cose]
cose)]+
(S.)23
5-3
-199time and setting the derivative
By differentiating with respect to
to zero,
S(t)) is minimized.
t
(and therefore
for which S2 (t)
the time t
we can determine
is given by:
S (v cosO - v
)
2 _2vvcos8
12
2
1
t* =+
5-4
*
5-4 is valid only for
Equation
Thus, we
t..
positive values oi
distinguish three cases.
First, if v 1 is less that
at t =0 and
that S..
v 2 , the minimum separation, S
Furthermore, according
is equal to S..
1
occurs
,
to 5-2, S
0
is less
v 1 is -less-than v ,-the--overall minimum separation
Thus,-when
12
occurs at the OM.
Second, if
v
> v2 > v cose
*
the minimum separation occurs again at
This time, however, S0 is
t.=0.
greater than S. so the overall minimum separation occurs when aircraft 1
is at P. and is given by 5-2.
Finally, if
v cose > v2
lt
*
the minimum
separation occurs at the
time t
given by
*
both aircraft are
5-1 we obtain
beyond P..
By substituting t
from
the formulas for the minimum separation
5-4, i.e.
when
5-4 into 5-3 and
S.
min
and for the
-200approach.
closest
of
time
the
at
when
coordinates
aircraft
Specifically, the minimum separation is given by:
S.
= S(t )
mi~n
S.sinO
5-5
1
1 + r2 -2rcosO
where, r is defined as the ratio v2/v1 of the aircraft speeds.
when a
slow aircraft (aircraft
When this happens,
to
achieve
aircraft
runway
separations
by slow
followed
utilization,
at
will
the
ones also
such pairs
case) follows a
2 in this
altitude separatinu
tight
the final approach
will arise during
Clearly, conflicts
will
will
takeoffs
OM.
be
scheduled
contribute to
occur less
ahead
since fast
Fortunately,
inefficiencies in
frequently
of
when the
aircraft were allowed
In addition,
random first-come-first-served sequence.
normally
faster one.
have to be imposed in order
runway schedule is optimized, than they would if
to land in their
leg only
the
following
slow
aircraft.
The
the optimal runway
fact that
have a smaller number of
is not
The
coincidental.
separations,
schedule will,
on the average,
fast-slow aircraft pairs in direct succession,
The same
however, that
principles are involved
have
to be
in both cases.
considered by
the runway
schedule, occur between the OM and the runway threshold.
Since altitude
there, fast-slow sequences
result in idle
separation is not an option
-201runway time.
The formulas for the
separation requirements between the
OM and the runway threshold are derived in Appendix A.
The
indicated
method
for
handling
fast-slow
sequences
maintain altitude separation
by clearing the fast aircraft
OM
and
keeping the
the slow
aircraft
crossing
higher.
The
altitude early
descent of
down to the
slow aircraft
can be
is to
1000 feet
timed so
that it
reaches the OM crossing altitude when the fast aircraft is at the OM and
the separation is So.
Table 5-1
illustrates the separations
this strategy.
In cases 1
through 6, v 2
and the timing
involved in
ranges from 140
to 90 knots
while the other parameters are kept constant at the following values:
S
0
= 3 nautical miles
d. = 2 nautical miles
v1 = 150 knots
6
Case
= 20 degrees
7 is the same as
case 6
except that S
is changed from
3 to 6
nautical miles.
In
addition
to the
value
of the
following values appear in table 5-1:
1.
v2: the speed of the slow aircraft,
minimum separation,
S. , the
min
TABLE 5-1
The effect of v 2 on Horizontal Separations
and on the Timing of the Descent
2
ta
Sa
(NM)
(NM)
(mins)
(NM)
2.00
0.00
4.87
3.29
3.04
2.80
3.53
0.55
6.14
3.35
2.65
1.86
3.37
4.50
0.91
6.73
3.51
2.43
2.11
2.07
3.71
5.00
1.09
6.80
3.64
2.12
100
1.82
2.13
3.93
5.13
1.14
6.55
3.80
1.83
90
1.56
2.07
4.08
5.00
1.09
6.12
4.01
1.58
90
3.69
3.84
7.84
7.50
2.06
11.76
6.00
4.31
x1
(NM)
2.09
1.45
2.44
110
5
6
CASE
v2
#
(knots)
Smin
(NM)
1
140
2.87
0.80
2
130
2.67
3
120
4
t1
t2
(mins) (mins)
*
.7
* S
0
=
6 Nautical miles for
case 7.
-2032. x,
y,, and x2 : the coordinates
of the two aircraft at the time
of closest approach,
3. t1:
the
interval between the
time of closest
approach and the
time of closest
approach and the
time aircraft 1 reaches the OM.
4. t2:
interval between the
the
time aircraft 2 reaches the OM.
the interval between the
5. ta:
time aircraft 2 starts its descent
and the time it (i.e. aircraft 2) reaches the OM.
6. S : the separation between the two aircraft at the time aircraft
2 starts its descent.
Note that--x- and-x--are now-relative to -the OM and not-to P. .
12
3.
*
that t1 is equal to t
Also note
since the analysis assumed that aircraft 1 was at
P. at time t=O.
I
The timing of the initiation of
descent by the second aircraft was
based on an assumed 500-600 feet per minute descent rate and an altitude
difference of
1000-1200 feet.
Accordingly, the
descent is initiated 2
minutes before the first aircraft reaches the OM.
By
comparing the
through 6,
the time
values of
the descent of
t
start
of closest approach.
t is 4.3 nautical miles.
a
while the
we note that,
the second aircraft
In
minutes after the closest approach,
time
and t ,
2a
aircraft are
in cases 3
starts approximately at
case 7, the descent
starts almost 2
and the corresponding separation at
In cases 1
and 2 the
still approaching
descent has to
each other.
In both
-204-
however, the
cases,
minimum approach
three nautical miles.
is almost
practical purposes, the aircraft have adequate
This means that, for all
longitudinal separation throughout.
table 5-1
From
between v
difference
that, as
we see
and v 2
increases.
formulate a strategy for the timing
the
of closest
time
decelerate near
result can
This
be used to
of the final deceleration which is
Namely. we can maximize the separation
going to take place on this leg.
at
decreases as the
expected. S
approach
the OM while the
allowing the
by
slow
aircraft to
fast aircraft decelerates immediately
after the turn to the final approach leg.
An extreme case of this would
t
be if the slow aircraft decelerates from
v2 when the fast aircraft is at the OM.
to its final speed
a speed v
Then the closest approach would
I
be the one applicable to the speed pair v1 . v 2
rather than the pair v1 ,
v2 '
have said earlier,
As we
used
to provide
Consequently
strategy.
aircraft
the timing of the
the required accuracy
it will not
always
decelerate
early
and
in the delivery
be possible
It is possible. however, to
slow
final deceleration is
at the runway.
to implement
the above
plan so that on the average fast
aircraft
decelerate
late.
In
particular, this can be achieved by appropriate selection of the desired
flight plans (see
section 5.3).
The selectioA of the
turn to the base
and final approach legs can also be used to achieve this goal.
-205guidelines that can
The general
the above analysis
be deduced by
are:
First,
the tight
separation, S
longitudinal separations to occur in
a
slow
aircraft is scheduled to
Consequently, altitude
safe
separations for
exist
in the
aircraft
will require smaller
the OM
the final approach region wnenever
land directly
separation will be
these fast-slow
runway schedule,
at least
at
,
behind a
required in order
sequences.
When
it is imperative to
1000 feet above
last few minutes of the approach.
faster one.
to assure
such sequences
maintain the slow
the OM crossing
altitude until the
Note the "laddering"' of aircraft that
will occur when n aircraft with speeds
v1
are
scheduled in
remain
2
>''
direct succession.
1000 feet
higher than the
than the third, etc.
will then be
> V
The operation
The
n
third aircraft
second, the fourth
will have to
1000 feet higher
on the base and final approacn legs
very similar to the operation of
a holding stack.
*As the
leader aircraft reaches the OM, the second aircraft is cleared to the OM
crossing altitude and all other aircraft
in the sequence are cleared to
1000 feet below their present altitude.
§S&corn,
should
be
Furthermore,
the
runway centerline intercept
close
if
the
to
the
OM
for
at
least
point (P. in
one
two aircraft are approaching from
of
the
figure 5-5)
aircraft.
a different side
of the runway, both should intercept as close to the OM as possible.
It
-206-
can
be
easily
the
increase
shown from
the
e between
early as
the
that
will etfectively
this
closest
time of
approach by
the slow
possible, while
to
down
should decelerate
its final
aircraft should
making the velocity
this will increase Smin by
Again
decelerate late.
at
5-5,
the two aircraft paths.
aircraft
fast
speed as
approach
Smin
separation.
increasing the angle
Thlird,
equation
difference between the two aircraft smaller.
We
the
since
conservative
section
this
conclude
time the
second starts its
larger.
In
nautical
miles
or
larger,
at
all
times,
separation
approach
glideslope
horizontal
fact, if S ,
speeds.
angle
is
approximately
separation
approximately 1100 feet.
separation
aircraft
independent of
reason for
The
aircraft, at the
descent along the glideslope,
the two
translate
this is,
3
to
their
will be much
at the
difference
of course,
3
OM,
is 3
maintain altitude
will
degrees.
an
this need
it reaches the O.
between the two
the required
Even
aircraft may intercept at
Instead, the second
strategy, the separations
Under this
for case 7).
and be on the glideslope when
its original altitude
OM crossing
glideslope at the OM,
though landings typically intercept the
the case.
the
be at
the OM (6 miles
altitude 3 miles before it reaches
not always be
will
second aircraft
calculations are
the
that
by noting
that since the
nautical
altitude
in final
miles of
difference
of
-20/-
5.7
Aircraft Guidance
all the aircraft within the TCA,
Having generated flight plans for
remains the
still
there
have
advanced navigation
them
to conform
airline fleet will
control systems
and flight
which will allow
plans with little
to 4-dimensional flight
it is
the future,
In
of the commercial
the great majority
expected that
guidance.
problem of
or no help
Consequently, we will concentrate here
from the automated ATM/C system.
respect to aircraft that do not have
on guidance issues that arise with
this capability.
basic problems
are two
There
first is the execution of turns
do with speed
has to
control.
that we
need to
from one leg to
do
Namely,
investigate.
the next.
we have to
The
The second
exercise strict
speed control throughout the approach, or can we allow the aircraft crew
some flexibility in controlling the speed?
Recovery from Errors in the Execution of Turns
5.7.1
us
Let
consider the
aircraft currently located at point P.
for
a turn
moment,
to a
that we
the airspeed, v ,
different leg
The aircraft's flight plan calls
at point
of the aircraft, we can
time to initiate
Since
shows an
Figure 5-6
P1 .
If we assume,
have perfect information on the wind
required to track the new leg.
turn, the
a turn.
execution of
for the
vector, W, on the
calculate the heading change
the aircraft has a finite rate of
the turn is 'also of interest.
Assuming a
-208-
Heading change
v
P
lPt
w
aVV
Figure 5-6.
The Kinematics of Aircraft Turns
-209standard rate
of turn, the actual
circle (shown
dashed in figure
point
the
Pt (and
The heading
initiated.
values
by the
required
therefore, determine
We can,
5-6).
the timing)
therefore
the turn
where
Processor for
has
to be
turn are the
to initiate the
change and time
Command
the arc of a
aircraft track will be
guiding conventionally
equipped aircraft along their assigned track.
When surveillance errors
as well as errors in
the estimates of W,
and va are introduced, inaccuracies will result in two ways:
Fir.st,
the required
heading change and
turn can no longer be accurately
will
generally be
turn.
Lateral deviations
early or late
new
out of
determined.
conformance with
from the
vector will not
to initiate the
As a result, the aircraft
its flight
intended track
These may increase
turn initiation.
ground track
the time
be parallel with
plan after the
will be
caused by
with time since the
the intended path.
Longitudinal deviations will also be present and will generally increase
with time.
Second,. during the turn, the
by the
to
quality of the tracking data produced
surveillance and tracking algorithms
the position
flying along
estimates that can
a straight line.
are significantly inferior
be obtained while
the aircraft is
Consequently, ,the accuracy in executing
the turn cannot be ascertained until good quality tracking data is again
available.
required
Typically. 30-40
after
the aircraft
seconds (i.e
comes out
8-10 radar
of the
"hits") will be
turn for
the tracking
-210errors to be reduced to the level expected for linear flight.
miles at the typical
to approximately 3-5 nautical
interval translates
This time
terminal area airspeeds for jet aircraft.
The degradation of tracking quality during a turn, requires all the
to recover from the
be able
is initiated.
next turn
particular importance
restriction is of
This
typicaily of the order
at
accuracy
approach.
the
of
stage
t-at
airport) provides better
located at the
(assumed to be
radar antenna
turn, before the
Fortunately, the proximity of the aircraft to the
of 5 miles in length.
ackming
errors introduced by one
and final approach legs, which are
for the base
miles. in order to
plan to be at least 3-5 nautical
legs of the flight
It is clear from the above discussion that, after a turn either the
path or
flight
flight plan
require correction.
aircraft will
of the
Large lateral errors can be corrected by the Conformance Resolurion Path
change which
heading
will
path
not
be the
one
plans
will have
take the
originally
to be
form of
closer to
intended by
the
any conflicts.
generated for
a small
its assigned
given, however, the
correction has been
Normally this deviation will not create
flight
will
the.aircraft
will bring
after the heading
Even
path.
the correction
Typically
Generator.
flight plan.
If it does, new
the conflicting
pair of
aircraft.
Small residual deviations from the
the
Command
Processor
prior
to
a
flight plan can be corrected by
subsequent
turn.
Longitudinal
-211-
deviations can be corrected by adjusting
if one
is specified
adjustments
conflict.
do
in
not need
They
have
to
the
the time for the speed change,
flight
correction,
be
taken
plan
for
as long
into
this
leg.
Lateral
as they do not create a
consideration,
however, in
calculating the heading change and the timing of a subsequent turn.
if the deviations
Finally,
cannot
be corrected
by
the Command
Processor, new flight plans will have to be generated.
5.7.2
Speed Control
By
necessity,
4-dimensional flight
changes at various time-points.
speed
needs
to
be actively
methodology
uses
controlling
timing at
ground
Nevertheless,
controlled
track
means
reduction at
that, even
Active
From
though the
some time-point, the
have to
specify speed
this does not impIy that
at all
selection
the runway.
when it is absolutely necessary.
simply
plans
as
times.
the
The proposed
primary
speed control
means
of
is only used
an operational standpoint, this
flight plan
may specify
Command Activator does
a speed
not transmit
that command to the aircraft.
An
example
Assume that an
will help
illustrate
aircraft is at point P of
how this
can
be accomplished.
the downwind leg (figure 5-7)
and that its current flight plan requires the aircraft to fly the ground
track specified by
the sequence of points P, P1 ,
reductions at points S1 and
S2.
P 2 ' B 3 , OM with speed
Furthermore, suppose that the aircraft
-212-
T
'P2
P3
p'
/'
Intended
Deceleration
Point
P
Deceleration
Point
Earliest
&S
Deceleration
Point
Figure 5-7,
Recovery From Early Speed Reduction
-213-
aircraft
The only
generated.
with
out of
to fall
the
The deceleration will
at point P.
decelerates unexpectedly
conformance and
way to bring
current flight
alert will be
a conformance
back into conformance
the aircraft
will be
plan
to
cause the
command a
speed increase.
Instead, the flight plan is corrected by
Clearly this is not desirable.
assigning a new base leg, thus shortening the length the aircraft has to
travel to reach the runway.
The
new flight plan, therefore, be the one
depicted in the dashed line.
There are
limits to how
early the pilot can
reduce the airspeed.
If the speed reduction occurs too early the aircraft will not be able to
reach the runway
path
on time. even if the
is assigned.
advance
the
Such situations
point- prior to
which
compensate for a speed reduction.
figure
5-7.
S
will depend
shortest available 2-dimensional
can be
determining in
avoided by
the flight
plan
generator cannot
As an example this may be point S0 in
on
the
assigned base leg from the OM, as
distance, d, of
the currently
well as the range of allowable values
for that distance.
The
general
summarized
method
as follows:
for
At the
approach path, the point S
compensate
advised
for
to
reaches S
0
,
avoiding
,
beyond
speed reductions,
maintain his
time of
speed
the pilot is free to
active
speed
control
can be
to a
new leg
of the
a turn
which the flight plan generator can
is determined.
until that
point.
The
Once
pilot
is then
the aircraft
reduce his speed at his discretion and
-214-
the flight plan generator will compensate
P
for it by adjusting the point
at which the aircraft will turn to base.
Note
that
maintains its
the
same
method
current speed longer
can also
be
used
than specified in
if
the aircraft
the flight plan.
Finally, as we have mentioned earlier, this method cannot be used on the
final
approach
leg.
The
final
speed reduction
controlled to achieve accurate delivery at the OM.
has to
be precisely
-215-
CHAPTER 6
CONCLUSIONS AND RECOMMENDATIONS
6.1
Summary
The purpose of
this research was to study
in the terminal area Air
decision making process
Traffic Management and Control (ATM/C) system,
and develop the methodology for the design and implementation of a fully
automated terminal area ATM/C system.
decision
making" methodology
Namely, the
Our work adopts the
which is practiced at
decision-making authority rests with
centralized
the present time.
the ATM/C controller.
Accordingly, the main focus of our work is the Ground Control Subsystem,
(AGCS), of the terminal area
aircraft
control
ATM/C system.
subsystem, and
the
Other subsystems (e.g.
various data
linK
the
systems) are
considered to the extent that they affect the decision-making process in
the ACGS.
In summary, the accomplishments of this research were in 4 areas:
First, we
studied the Automated Ground
at a top level of
functional
Control Subsystem, (AGCS),
aggregation, identified its elements, described their
requirements,
and studied
their
interrelationships.
From
this functional description of the AGCS,
emerged the concept as well as
the
automated terminal
top level
software design
for the
area ATM/C
system, which provided the framework for the remainaer of our research.
-216-
Second.
developed a
we
heuristic
algorithm
for
algorithm
is based
on
The
uses the
The
proposed.
he
that
methodology
Shifting
Position
Constrained
and
[DEA 76],
Dear,
of
work
schedules.
runway
generating efficient
the
with the
we designed and implemented a
Finally,
remaining automation functions.
as
well
controller- as
ATM/C
the
with
scheduling
runway
interactions of
detail the
in
we examined
Furthermore,
complexity.
to its
contribute
factors that
the
and examined
problem
scheduling
the runway
of
general formulation
very
aircraft on multiple runways and was
algorithm is capable of scheduling
specifically designed to be implemented in a real-time environment.
developed a
we
Third,
4-dimensional
flight plans.
flight
that can
plans
changing
runway
flight
planning
In
schedule.
methodology
planning
on
modified to
adapt to
connection with
this
of aircraft guidance
can be
the
is designed
Our methodology
be easily
the problem
discussed
generating conflict-free,
methodology for
of
part
the
a dynamically
methodology, we
how our flight
and showed
precise 4-dimensional
to reconcile
used
to generate
automated
system, with
ATM/C
conventional navigation capability onboard the aircraft.
Fourth.
we
facility
simulation
development
and
software is based
system.
"critical
to
be used
testing of
as
a testbed
automation
of
is a
software
vast area
terminal area
real-time
further research,
for
software.
on our concept for the
ATM/C automation
mass"
implemented a
and
designed
The
design
of the
automated terminal area ATM/C
of research
before experimentation
and requires a
and
testing can
-217begin.
We feel that this work has significantly contributed towards the
achievement
specifically
automation
of
this
critical
mass.
Furthermore,
the
designed
to
facilitate
functions.
In
particular, the simulation
further
software was
development
of
ATM/C
facility has the
following characteristics:
1. Modular software design which
allows easy moaification of
specific simulation functions with
others.
Monitor
In
particular,
process
processes.
This
be simulated.
human
control any
necessary
direct control of
of realism in
particularly
of
them
flexibility
in
simulation.
For
may
modifying
of
of subordinate
stations may be
stations will
"human
be
factors"
add an
may prove
research.
loops and the fact that any
implemented,
the
flow
of
allows
events
great
in
the
surveillance data
four seconds J= AU aircraft.
which is specifically
the surveillance
positions to
the simulated aircraft
example, tracking and
algorithms, this not may
Simulation
the simulation and
for
are now updated once every
In research
the
number
Pseudo-pilot
Finally, the concept of timer
number
of
allows multiple controller
operators.
extra degree
logic
In addition, pseudo-pilot
added, allowing
by
can
the
little or no effect on
interested in tracking
be satisfactory since the timing
data, which depends
bearing may be of importance.
on the aircraft
The simulation logic allows
-218-
such
be
to
modifications
relatively little
with
made
effort.
is based on detailea models
2. The simulated aircraft motion
and
detail.
of
instrument
cockpit
flight
aircraft
Air Traffic
aircraft
quality
by
seen
aircraft
of
the
ATM/C
flight
is
of advanced terminal
Control since the
precision with which
guided along
can be
as
in the study
particularly important
area
moaelled in
very realistic representation
modelling
Proper
controller.
also
are
readings
results in a
This
Errors in the navigation
system.
of the aircraft control
4-dimensional flight plans
has tremendous effect on the flight planning logic.
3. The
interface between
functions
designed
specifically
was
and
experimentation
and the automation
the controller
testing
of
to
alternative
allow
display
concepts. This was done by structuring the software in two
"layers".
The lower
the screen and
displayed.
layer manages the
proper update of
is independent of the actual
data that is
Thus the displayed data as well as the display
format can be
changed by modifying only the
the I/0 interface.
top layer of
-2196.2
Recommendations for Future Research
main areas
are two
There
and development is
where research
still needed before a successful implementation of an automated terminal
area ATM/C system is accomplished:
1. Algorithms for Traffic Flight Plan Generation
program has shown that automatic
The Metering and Spacing
be achieved.
flight plan generation can
version, however, has significant
out (chapter
pointed
may
software
incompatible
totally
be
scheduling function
which is expected to
methodology
in this
proposed
for
the runway
with
provide most of
current ATM/C system capability.
the improvement over the
The
drawbacks which we have
importantly, the current
Most
1).
The current test
flight
automated
work should
provide a
generation
plan
solid basis for
further work in this area.
our
From
experience
flight
plan
ground
for
methods.
be
a
with
the subject
will be
generation
applications
of
a
we
believe that
particularly fertile
Artificial
Intelligence
Specifically, the conflict resolution
"learning
regarding the
program"
which
acpumulates
selection of alternative
conflicts are identified.
logic can
experience
flight plans when
-220-
2. Human Factors research and experimentation
The introduction of automated runway scheduling and flight
substantially increase
generation will
plan
available
information
of
amount
large
the already
to
the
ATM/C
This creates first the need for experimenting
controller.
The interaction
with alternative methods of presentation.
of the ATM/C controller with the automation software is of
re-evaluate
the
of
allocation
factors
human
chapter 4 the need to
Finally, we discussed in
research.
in
area
significant
other
the
course
responsibilities
among
various controller positions in the terminal area.
Independent of our views on
an
automated ATM/C
the role of the controller in
environment, we
believe that
in all
three of the above areas there is a need for quick testing
and evaluation of alternative concepts.
an
effective
concept
level
tool
for
due to
TASIM can provide
research
factors
human
the relative
ease with
at the
which the
display format and the I/0 interface can be modified.
The existence of a precise runway
all
aircraft
in the
terminal
schedule and of flight plans for
area give
rise
to the
possibility of
improving the capabilities of a number of "establisned" functions in the
ATM/C system.
In chapter 2 we described possible improvements in hazard
detection and tracking through algorithms
as surveillance
data.
which use flight plan as well
Additional improvements may
be possible in both
-221-
these
functions
equipment
(e.g.
This can be
if
them
we provide
with readings
speed and vertical
heading,
achieved through the use of
from
the onboard
speed indicators, etc.).
the air-to-ground digital data
link for mode-S equipped aircraft.
Finally
4 we
in chapter
scheduling and congestion
discussed the
management or flow control both
fixes and at the departure
gates.
used to allow much more efficient
The runway
schedule is therefore
automated enroute flow
connection between runway
Runway scheduling information can be
flow control in the enroute airspace.
a valuable source
control systems such as the
concept description of AERA.
at the entry
ot information for
one included in the
-
blank page -
-222APPENDIX A
INTERARRIVAL DYNAMICS
This
appendix presents
the mathematical
the minimum time separation between
runway.
formulas for determining
consecutive arrivals using the same
The calculations assume that the two basic ATC rules related to
the runway operation are:
1. No two
aircraft are permitted
on the same
runway at the
same time.
2.
Coaltitudinal aircraft under
ground control must maintain
a specified horizontal separation.
It is also
same
runway fly
equal
to
the
assumed that all
a common
aircraft's
controlled aircraft
final approach
preferred
path at
approach
arriving at the
a constant velocity
speed.
The
preferred
approach speed depends upon such parameters as the type of aircraft, the
landing
preferred
weight, the
weather conditions,
approach speed
arrives at the entry fix.
is specified by
and pilot
the pilot
preferences.
This
when the aircraft
Consequently. two identical aircraft may have
different preferred approach speeds.
The minimum
time separation at
landings is analytically determined as
the runway between
two successive
a function of the final approach
-223-
length.
and
speeds
approach
the
the
minimum
horizontal separation
Specifically, let:
distance.
v land (i) = the approach speed of aircraft i
tocc (i) = the runway occupancy time of aircraft i
= the
s..
ij1
minimum horizontal
followed by aircraft
separation for
j
= the length of the common final approach path.
F
Then, t. .(s. .F),
iJ
the minimum time separation at ,the runway between
iJ
the landing of aircraft i followed by aircraft
max[t
aircraft i
j is
given by:
C(i) ; s.
./v (j)]
13
land
occ
when v land(i) is less than v land(j). and by:
max[t occ(i) ; Sij/vl
(j)
(ladj)
- 1/vland(i))]
+ F(1/v
when v land(i) is greater than v landj).
To simplify this expression, let c
then,
be defined as follows:
0
vland (i)<v land Q)
- 1/v and (i)
1/v land(j)
/
ladlad
(j)
V land (i) -> v landCJ
-224t..(s..,F) = max Et
1j
1J
Figures
; s. ./v
(i)
ij
land
A-2 illustrate
A-1 and
occupancy
runway
occ
time
aircraft
of
(j) + Fe..]
13
landing siLuations.
the two
1
less than
be
to
is assumed
The
s ./V land (2) in both cases.
In
the overtaking
situation of figure
than aircraft 1 and consequently,
2 is faster
the point of closest approach between
the two aircraft occurs when aircraft
runway.
A-i, aircraft
the first one touches down on the
This closest approach equals s12/V land (2).
In
the
aircraft
opening case
1 and
the point
begins its final approach.
beginning
the final
time units
of
figure A-2,
of closest
aircraft
2 is slower than
approach occurs
when aircraft 1
Aircraft 1 lands F/vland(1) time units after
approach and aircraft
after aircraft 1
2 lands
(F + s12 Vand(2)
begins its final approach.
Thus the time
between the two landings is:
(F + s1 2)/v land (2) = s /V land (2)
+ Fe1 2
12
Suppose now that three aircraft, with approach speeds 120, 135, and
150 knots respectively, have identical preferred times of arrival at the
runway,
(arbitrarily set
to 0).
Table
A-1 presents
the minimum time
separation for all possible combinations of successive arrival pairs.
final
approach
length
of
5
nautical
miles
was
assumed
for
A
the
-225-
ACs
2
AC 2
N
t 12 = s12 /Vland (2)
Figure A-i.
Minimum Interarrival Separation (Overtaking Case)
AC 1
F
S12 + F
t12
Figure A-2.
=
Vland kz)
~
land (1)
Minimum Interarrival Separation (Opening Case)
-226-
TABLE A-1
Minimum Interarrival Ti me Separation (Seconds)
At the Runway
Leading
Aircraft
(Knots)
Following Aircraft (Knots)
. 120
135
106.67
150
120.00
135
150
80.00
72.00
72.00
93.33
S..
F
= 3 NM
= 5 NM
-227-
In
calculations.
this
the
example,
occupancy times
runway
are
irrelevant as long as they do not exceed 72 seconds.
There are six sequences in which
the assigned
lists
each
to the
landing times (rounded
and the average
landing sequence
The delay
aircraft.
the aircraft can land.
Table A-2
nearest second) for
delay experienced
by the three
to be the difference
for each aircraft is assumed
and the preferred landing time (which
between the assigned landing time
is zero in this case).
even
in
this
in this example.
points are brought out
Several important
situation, the
simple
sequence
affects the runway utilization as demonstrated
the
same
landing
time
situations
consider
schedule.
Second, the
average delay
(e.g.
occur
more
(e.g.
than one
often
by the range in the time
fact that some sequences have almost
sequences
sequences 1
very
operations greatly
of
200) and the variation in the average
the last aircraft lands (152 vs.
delay (74 vs 106.67).
First,
and
3
and 6)
2) is not
and indicate
efficiency measure
that
in
or the
same last
coincidental.
it
Such
is necessary to
optimizing
the runway
TABLE it-2
A Comparison of Lar ding Sequences
Landing
Time
Seq. 3
Landing
Time
Seq. I
Landing
Time
1
0
1
0
2
0
2
80
3
72
1
107
3
152
2
165
3
179
Average
Delay:
74
Average
Delay:
Average
Delay:
75.67
95.33
Landing
Time.
Landing
Time
Seq. 4
Landing
Time
2
0
3
0
3
0
3
72
1
120
2
93
1
192
2
200
1
200
Delay:
.
Average
Average
8g
Delay:
106.67
Average
Delay:
97.67
-
blank page -
-229APPENDIX B
THE KINEMATICS OF THE INTERCEPT VECTOR
The kinematics of determining a vector
moving target can be expressed as
a function of the relative speeds and
the relative position of the aircraft
discussion, turning radii, the time to
will be
ignored.
to cause the intercept of a
To simplify this
and the target.
turn and the time to decelerate
These complicate unnecessarily
the basic concept and
can be accounted for by calling turns and speed changes earlier tnan tne
idealized model requires.
Consider figure B-1.
and is moving along the x axis
at a constant speed vb*
at a -point (x0,y0)
and has a ground speed v .
to
approach speed
the aircraft's
is at the origin
At time t=0 the target box
va
The
This
The aircraft is
speed may be equal
intercept problems
can be
stated as:
1.
Is it possible to intercept the box?
2. What heading e is required to intercept?
3. What
is the
time tv
to call the
speed change
if v0 is
greater than va?
4. What is the time to intercept. t.?
Since it is obvious that the required heading 0 depends on the time
the
speed change
is called,
change at t = t./2.
i
v
we can
The average
set the
time to
call that speed
aircraft ground speed in the interval
(x ,yO)
/
/
Figure B-1.
The Kinematics of the Intercept Vector
-231(t , t.)
o
I.
will
in
flexibility
v = (v + v )/2.
o
a
be
correcting
a
This
turn that
was
assumption
not
allows
perfectly
some
timed by
change after the turn is completed and
adjusting the time for the speed
good tracking of the aircraft has resumed.
Accordingly, we can
assume that the aircraft travels
at a speed v
until it intercepts the box at time ti.
from time t
The equations of motion are:
r
= it
= vcose
x = x + vtcosO
0
= vsinO
V= v
- vtsine
At the time.of intercept, y = 0 and x = r.
=t..
Thus,
1.
3.
y
o
= vt.sine
1
it. = X
+ vt.cosO
From B-1, the time of intercept is given by:
t. =
I
vsinQ
From B-2, substituting for t .
r
-
vcose
vsinO
x
0
Y
B-1
B-2
-232x
0
cose +v
yO
From B-3, if we know the
e,we
angle
B-3
sine
can solve for the
ground speed ratio r/v, and the intercept
of interest and not the actual values of x
a = tan
ratio r/v and
of
e,the
and yo.
box
from the box, for any given values
can be intercepted as
and the point
of intercept on the
long as the
The time to interceDt the
aircraft is on the dotted line of figure B-1.
box t.,
If we denote by
-
the relative bearing of the aircraft
of the
Note that only the ratio is
ratio x /y .
x axis, will
depend on the
position of the aircraft along the dotted line.
The relative bearing.
, determines the position of the wand on the
final vectoring display discussed in section 4.11.
-233-
REFERENCES
AHO 74
V., Hopcroft,
Aho, A.
and Ullman, J.D.,
J.E.,
Ihe Design anjd
L Computer Algorithms, Addison-Welsey,
Analysis_
Menlo Park,
California, June 1974.
ATH 71
Athans, M.
and Porter, L.
Scheduling
and Holding
1971
Automatic
Joint
W., "An Approach to Semi-Automated
Air Traffic Controlu
Strategies for
Control
Conference
P reprints.a August
11-13, 1971.
BLU 60
Blumstein, A., An Analytic
Ph.D.
BRA 77
Inv;estigation af AirporL Capacity.
Thesis, Cornell University, 1960.
S. P.,
Bradley.
Hax, A. C.,
and
Magnanti, T. L.,
Applied
Mathematical Programming, Addison-Wesley, Reading, Ma., 1977
COU 78
Couluris, G.
Impacts
AIM
Qf
Report No.
J., Tashker, M.
Automation:
G., and Penick, M.
Human
FAA-AVP-78-1, DOT-FAA,
C., Policy
Factors Considerations,
Office of Aviation Policy,
Washington, D.C., January 1978.
CRE 73
Credeur,
L.
et al.,
Management Systems
NASA, Langley
1973.
Study
UIla
the.
Qn Advanced Aircraft
Research Center, CR-132278,
2I
Qf
Air Traffic
and A-vionic Systems,
Hampton, Virginia,
-234-
DEA 76
Terminal
Report
FTL
Area,
R76-9,
Laboratory. M.I.T., Cambridge, Ma,
DRA 67
Drake,
A.
in th. Near.
The Dynamic Scheduling af Aircraft
Dear, R. G.,
W., Fundamentals
Transportation
Flight
August 1976.
Probability Theory,
Qf Applied
McGraw-Hill, New York, 1967.
DUN 75
Dunlay, W. J. Jr., "Analytic Models of Preceived Air Traffic
Conflicts," in
Control
Vol.
9, No.
149-164.,
2, May 1975, pp.
DUR 77
Transportation Science.,
Durocher, C. L., Ground Controlled Precision Landing Delivery
in
Presence
The
Radar Disturbances,
2f
Report R77-2,
FTL
Laboratory, M.I.T., Cambridge,
Flight Transportation
Ma, May
1977.
GOL 81
Goldmuntz, L., et al., Automated En-Route Air Traffic Control,
Report #
FAA-EM-81-3, FAA, Office of
Systems Engineering and
Management, Washington D.C.,March 1981.
HEI 76
Heinz,
V., Stochastic
S.M.
Thesis,
M.I.T.
Simulationa QL Terminal Area Airspace,
Department
of
Aeronautics
and
Astronautics, Cambridge, Ma, September 1976.
HOC 74
Hockaday,
Capacity
171-180,
S.,
and
Kanafani
Analysis", in
1974.
A.,
"Developments
Trasportation Research,
in
Vol.
Airport
8, pp.
-235-
HOF 72
W.
Hoffman,
Error
Functional
Hollister,
C.,
Analysis
and
W.M.,
R.W.,
Evaluation,
Consepts
AI
for
Simpson,
DOT-TSC-212-72-1, Aerospace Systems Inc., Burlington, MA., May
1972.
HOL 74
ATC
V., "Geneology of Terminal
C., and Garceau, T.
Holland, F.
Automation, 1
MITRE
2, The
Revision
Corporation, March
1974.
HSI 76
Management and Control, FTL
Aix Iraffic
Terminal Area
Study if Advanced
C., An Analytical
Hsin, C.
Report R76-13 Flight
Transportation Laboratory, Cambridge, Ia, September 1976.
MAT 77
System Software Specification, Marine
System (MATCALS),
Landing
Air Traffic Control and
Vallejo, Ca,
Systems Engineering,
September 1977.
ODO 69
Odoni,
Analytical
R., An
A.
Air- Traffic
Investigation QL
Control in .t.h Vicinity. Af Terminal Areas, M.I.T.
Research Center, Techical Report No.
PAP 82
H.,
Papadimitriou,
C.
Optimization:
Algorithms
and
and
Operations
46, December 1969.
Steiglitz,
K.,
Complexityi,
Combinatorial
Prentice-Hall,
Englewood, N.J., 1982.
PAR 79
Pararas,
Facility,
J.
D.,
M.I.T.
Sequencing
Flight
78-2, September 1978.
Tasks
2n
a
Multi-Procesra
Transportation Laboratory, FLT-Memo
-236-
PAR 81
Development
Traffic
DL
& Terminal
and
ConLrol.
W.,
R.
Simpson,
D.,
J.
Pararas,
Hoffman,
and
Area Simulation
Landing
System,
f=r
W.
C.,
Marine Ai
Flight Transportation
Associates, Interim Technical Report FTA-TR-81-201, Cambridge.
Ma. January 1981.
PAR 81a
Pararas,
J.
D.,
Airport
An
Associates,
Transportation
Model,
Capacity
Flight
Technical
Report
FTA-TR-117.
Computer
PrograM
Development
Cambridge, Massachusetts, 1981.
PAR 82
Pararas,
J.
f=r
Specification
et
D.,
al,
jj= AI.M
Subsystem
POR 79
Poritzky,
M.I.T.,
Cambridge,
FTL Report R82-1,
Flight Transportation Laboratory,
Massachusetts,
Df .tha MVSRF.
January 1892.
S.B.,
Decisionmaking",
"ATC Automation:
Presented
at
the Dilemma
ATCA,
Atlantic
of Automated
City.
N.J.,
October 1979.
PSA 78
Psaraftis,
Specially
R.
Structured
Transportation.
Dynamic
N.,
Segnencing
Ph.D.
Programming
and
Dissertation.
Algorithms
Routing
Problem
Department
of
f=r
in
Ocean
Engineering. M.I.T., Cambridge. Ma, September 1978.
RAT 64
Ratcliffe,
S..
"Congestion
Navigation Journal. Vol.
in Terminal
Areas," Institute Q.
17, April 1964, pp.
182-186.
-237ROW 78
"Terminal
G.,
Rowland,
Conflict
August 1978.
Report, DOT-FAA, Washington, D.C.
SAR 71
Sarris, A.
Air
H.,
and Athans,
Iraffic Control,
SRDS Progress
In
Alert,"
M., An Algoritbm f=r Terminal Arjea
Systems Laboratory,
M.I.T., Electronic
Report ESL-P-466, December 1971.
SCH 73
D.
Approach
to Terminal
Aircraft, Vol.
SIM 64
and
K.,
Schmidt,
R.
Simpson,
Operations
Area Air
10, No.
W.,
Swain, R.
L.,
"An
Traffic Control,"
3, March 1973, pp.
Terminal
Area,
Ph.D.
JQurnal at
181-188.
An Analytical Investigation
in jhe
Optimal Control
af Air Traffic
Dissertation,
Department of Aeronautics and Astronautics, M.I.T., Cambridge,
Ma, May 1964.
SIM 66
Simonnard,
R., Linear
Programingi,
Prentice
Hall, Englewood
Cliffs, N.J., 1966.
TAL 78
Talley, J. R., "Basic Metering
and Spacing for ARTS III," In
SRDS Progress Report, DOT-FAA, Washington, D.C., August 1978.
TAL 80
Talley, J. R., "Basic Metering
and Spacing for ARTS III," In
SRDS Progress Report, DOT-FAA, Washington, D.C., 1980.
WES 81
Wesson,
R.,
et al,
Scenario- fa
Control, Prepared for the FAA
Evolution Zf
Air Traffic
by the Rand Corporation, Report
No R-2698-FAA, Santa Monica, California, November 1981.