FLIGHT TRANSPORTATION LABORATORY
REPORT R 91-4
PRESENTATIONS FROM THE
MIT/INDUSTRY COOPERATIVE RESEARCH
PROGRAM ANNUAL MEETING, 1991
DEPARTMENT
OF
AERONAUTICS
Belobaba, Williamson, et al.
ASTRONAUTICS
May 1991
FLIGHT TRANSPORTATION
LABORATORY
Cambridge, Mass. 02139
FTL REPORT R 91-4
PRESENTATIONS FROM THE MIT/INDUSTRY COOPERATIVE RESEARCH
PROGRAM ANNUAL MEETING, 1991
TABLE OF CONTENTS
Origin-Destination/Segment Seat Inventory Control: Modeling and
Implementation Issues.
Peter Belobaba
1
Application of Network Solutions to Origin-Destination Seat Inventory Control.
14
Elizabeth Williamson
Planning and Scheduling Tasks in a Dynamic Environment.
Lyman Hazelton
46
Concentration in U.S. Air Transportation: An Analysis of Origin-Destination
Markets since Deregulation.
73
Jan van Acker
Pricing in the Airline Industry: Current Practice and Future Research.
Theodore C. Botimer
102
Changes in Origin-Destination Passenger Traffic Flows: Newark Airport.
Chung Mak and Peter Belobaba
129
Airline Seat Inventory Control for Group Passenger Demand.
Peter Belobaba and Tom Svrcek
157
Airline Crew Scheduling Re-visited.
Robert W. Simpson
173
Scheduling Systems: Computer Aids for Execution Rescheduling.
Dennis F.X. Mathaisel
187
-1-
O-D / SEGMENT
Modeling
SEAT INVENTORY
and
CONTROL:
Implementation Issues
Professor Peter P. Belobaba
MIT Flight Transportation Laboratory
MIT/Industry
Cooperative Research
Annual Meeting
May 23, 1991
Program
-2-2-
PASSENGER
e
ITINERAR Y CONTROL
In contra! to the control of seat inventories
by flight leg/booking class, passenger
itinerary control requires methods that can
distinguish among Dassenaer itineraries
itineraries
vying for seats on the same flight
leg, even
within the same fare product "category"
Passenger Itinerary Control concepts can
be applied through:
DUAL/OVERLAP FLIGHTS: Two or more flight
numbers representing different itinerary
"paths" a ssigned to a single flight leg.
POINT OF SALE CONTROL: Seat availability
is differentiated between points of sale due
to currency or net revenue differences.
SEGMENT CONTROL: Seat availability
managed by booking class and passenger
itinerary on a multiple-leg flight with the
same flight number.
O-D CONTROL: Seat availability
managed by fare category and O-D
itinerary, even across connecting flights.
-3-
PASSENGER
)VERLAP
DUAL /
D7W
ITINERARY
CONTROL
FLIGHTS:
NWoL09
L4W
tW1b34l
POINT OF SALE:
KL 016
AMs
RB
SEGMENT
CONTROL:
CONNECTING
O-D
AC 113
CONTROL:
2
0
DL145
liou
00s
SAV
-4-
NETWORK
OPTIMIZATION
APPROACHES
Traditional O.R. approach to dealing with
itinerary control is to perform a joint
optimization over the entire network of flight
legs defined by the problem.
.
Network formulations and/or mathematical
programming approaches are used to find
the optimal a Illocation of seats to each
origin-destination itinerary and fare type
(ODF) on each flight leg:
--
demand forecasts and fare
values for each ODF
requires
representation can be
deterministic or probabilistic
--
problem
--
solution
--
ODF allocations are "optimal" given
ensures balanced ODF
allocations across flight legs
assumed mathematical formulation of
problem
-5-
9
Implementation of these "optimal" network
ODF seat allocations as ODF booking limits,
however, can have substantial negative
revenue impacts.
--
use of partitioned or discrete booking
limits lowers expected revenue relative
to nested limits
revenue impact becomes
larger with more ODF allocations
--
negative
--
dynamic
simulations show that use of
"optimal" partitioned segment/class
limits on a 3-leg flight can result in
revenue reductions of 1% to 2% relative
to simpleh leg/booking class control
The oDtimal solution to an assumed
mathematical formulation does not
maximize revenues when implemented this
way.
-6-
NESTING
OF NETWORK
SOLUTIONS
OPTIMAL
Nestina of ODF Allocations Within Each
O-D/Seament
0
Network ODF allocations for each
O-D/segment are nested into a shared
inventory of seats, in order of normal nested
booking classes
--
each ODF allocation is treated as a fare
class "protectic n level" within the
O-D/segment "nest"
--
each flight leg still has a discrete
allocation of s eats for each
segment/O-D itinerary
--
summing discrete allocations
gives sub-optimal nested booking limits
within each O-D/segment nest
simply
-7-
EXAMPLE: 2 segments on Leg A-B
(Capacity = 100)
ODF
Optimal
Allocation
Segment
Nesting
YAB
8
42
BAB
12
34
MAB
06
22
QAB
16
16
YAC
18
58
BAC
7
40
MAC
23
33
QAC
10
10
-8-
2. Joint Allocation and Nestina Within
O-D/Segment
*
Published by Curry in Transportation Science
(1990)
--
jointly finds optimal allocation
to each "O-D nest" and nested limits on
each booking class within the nest
approach
solution to the formulated
problem
--
optimal
--
still a discrete allocation of seats to
each segment/O-D
(i.e., "O-D nest")
-
Both approaches above can have positive
revenue impacts compared to leg/booking
class control provided that number of
discrete O-D nests does not become large
e
Otherwise, negative revenue impacts can
result when implemented as a control
methodology
1NIMMINIM1111W
161
k1i
NNINSINN
hi
,11131
111111
11110
1
1j,11191M
1
Will,
-9-
Prices
of ODF Allocations on Shadow
Nestinga
3.
Shadow
Nestina
on
3.
*
Described in Williamson's (1988) MIT Master's
thesis
ODF allocations a re ranked and
nested in order of shadow price values
derived fro n the optimization algorithms
--
optimal
--
the shadow
--
on each flight leg, ODFs with highest
price of an ODF allocation
the amount by which total expected
network revenue will increase (or
decrease) if one additional seat is
allocated to that ODF
shadow prices values receiv e greatest
availability -- ODF allocations are
treated as "protection levels" for nesting
purposes
-10-
EXAMPLE:
0 DF
YAB
BAB
YAC
BAC
MAB
MAC
QAC
QAB
e
2 segments on Leg A-B
(Capacity = 100)
Optimal
Allocation
8
12
18
7
6
23
10
16
Shadow
Price
225
200
190
165
110
40
10
0
Nested
Limits
100
92
80
62
55
49
26
16
Implementation into control structure
possible through virtual inventorv classes
defined by shadow price ranges
Nesting of optimal ODF allocations on
current shadow prices results in theoretically
sub-optimal booking limits for different
ODFs
e
Yet, a large number of dynamic booking
simulations of this approach as a control
methodology show consistent and
substantial revenue improvement over
leg/booking class control
-ll-
Network
4.
4. Network
*
"Bid-Price"
Developed at MIT:
Williamson (1990)
Approaches
Simpson (1989) and
optimization can also produce
shadow prices on the capacity
constrain associated with each flight
leg, or "bid prices"
--
network
--
Bid Price Control of seat inventories
simply requires a comparison of the fare
of the requested ODF itinerary and the
sum of the bid prices involved in the
itinerary.
--
Implementation requires frequent (real-
time?) updating of network bid prices to
overcome absence of booking limit
controls
*
Bid Price Approach is a sub-optimal control
application of optimal network solution,
which over comes negative impacts of
discrete ODF allocations through "implicit
nesting" of availability through bid price
evaluation decision.
-12-
SUMMARY
*
OPTIMIZATION
APPROACHES
NETWORK
--
Network optimization approaches produce
optimal ODF seat allocations over a
network of flights:
--
require
ODF demand forecasts and
fares
--
*
partitioned ODF allocations
that must be "nested" in sub-optimal
ways to have positive revenue impacts
generate
Truly optimal solution for control of ODF
itineraries over a network requires an
approach that
for nesting of ODFs explicitly
--
accounts
--
allows "desirability"
--
recognizes
--
overcomes "small number" problems of
of ODFs to change
as demand materializes
dynamic nature of future
booking process
forecasting ODF demands
-13-
CONCLUSIONS
0
Network optimization methods produce an
optimal solution to an assumed
mathematical formulation of the
O-D/segment control problem
Implementation of optimal solutions in actual
reservatior is control structures can lead to
negative revenue impacts if done
incorrectly
*
Nesting of optimal network ODF allocations
for control purposes is a sub-optimal
solution, although some nesting approaches
consistently produce better revenue
impacts than others
* No one has formulated, let alone "solved"
the dynamic, nested ODF network seat
inventory control problem
-14-
APPLICATION OF NETWORK
SOLUTIONS TO
O-D SEAT INVENTORY CONTROL
Elizabeth L. Williamson
Flight Transportation Laboratory
Massachusetts Institute of Technology
Presented to
MIT/Industry Cooperative Research Program
May 23, 1991
Cambridge, MA
-t
-15-
Introduction
Reviewing network seat inventory control techniques and applying
them to three different multi-leg examples, using real airline data:
1)
2 Leg Flight
A
B
C
B
C
4 Fare Classes
3 OD Pairs
12 ODF Combinations
2)
3 Leg Flight
D
4 Fare Classes
6 OD Pairs
24 ODF Combinations
3)
4 Leg Flight
e- -
A
4 Fare Classes
10 OD Pairs
40 ODF Combinations
e-
B
-
+
C
-
+
--.
D
E
-16-
Network Solutions
Nested on Shadow Prices
- Network formulation used to find seat allocations for each
ODF over an entire network of flights.
- Distinct allocations are nested according to the shadow price
of each ODF.
*Shadow Price: The amount the optimal system revenue value
would change if one more seat was made available to the
given, ODF.
-17-
Nested Deterministic by Shadow Prices
3 Leg Example
Leg BC - Capacity=90
ODF
ACY
BCY
ADY
ACM
BCM
BDY
ACB
ACQ
BCB
ADM
BCQ
BDM
ADB
ADQ
BDB
BDQ
Seats
Allocated
2
10
3
1
22
6
4
14
12
1
15
0
0
0
0
0
Fare
519
440
582
344
315
440
262
231
223
379
197
307
302
269
Shadow
Price
Booking
Limit
322
243
216
147
118
90
88
78
74
52
46
65
34
26
75
74
42
13
28
16
0
15
-59
-64
221
-97
-145
199
-167
0
0
0
0
0
-18-
NETWORK SOL'NS NESTED ON SHADOW PRICES
2 Leg Flight
2.6
2.4
l
22
w
2
V
1.8
1.6
u
1.4
0
1.2
1
C
0.8
0.6
16-
0.4
S.
020
-02
-0.4
0.7
0.74
0.78
0
0.82
0.86
Load Factor
NDSP
+
NPSP
0.9
0.94
0.98
-19-
ON SHADOW PRICES
NETWORK SOL'NS NESTED
3 Leg Flight
4
(12
Li
N
3
2
U
61
2
0
S..
'.4
0.
61
U
63
S..
61
'.4
'-S
4.'
U
U
Li
U
a.
0.76
0.84
0.8
0
0.88
Load Factor
NPSP
+
NDSP
0.92
0.96
-20-
NETWORK SOL'NS NESTED ON SHADOW PRICES
4 Leg Flight
0.55
0.85
0.75
0.65
Load Factor
0
NDSP
+
NPSP
0.95
-21
DISTINCT DETERMINISTIC
APPROACH
o00
500
-L
400
300
200
100
0
0
5
10
15
20
25
30
35
DISTINCT PROBABILISTIC APPROACH
500
500
400
300
200
100
0
0
5
10
15
20
25
30
35
3
13
EMSR APPROACH
00
500
400
300
200
100
00
5
5
10
10
15
15
20
25
20
25
-
-22-
EXAMPLE
Single Leg, 4 Fare Classes
MEAN
20
15
30
25
Y
M
B
Q
STD
7
5
10
8
FARE
500
350
200
150
ALLOCATIONS
DETER
Y
M
B
20
15
30
25
Q
PROB
EMSR
27
19
31
23
17
20
27
36
BOOKING LIMITS
Y
M
B
Q
NDSP
NPSP
100
80
65
35
100
73
54
23
EMSR OPTIMAL
100
83
63
36
100
83
62
33
-23-
Initial Allocations
3 Leg Example
AB Leg - Capacity=75
Distinct
Probabilistic
Distinct
Deterministic:
AB
AC
AD
Y
M
B
Q
25
2
3
3
1
0
7
4
0
26
4
0
AB
AC
AD
Y
MB
Q
28
3
2
5
10
0
0
1
0
26
0
0
-24-
Difference in Allocations
(Prob - Deter)
Y
AB
AC
AD
l
1
MB
Q
1
0
0
-4
0
-3
0
-25-
Comparison of Allocations
Over 15 Revisions
Mean
AB Y
25.2
25.1
24.8
24.0
22.8
22.0
20.4
19.3
16.9
15.6
12.3
9.2
8.6
5.9
2.6
Deter
Alloc
Prob
Alloc
25
25
25
24
23
22
20
19
17
16
12
9
28
28
28
28
28
26
26
26
25
23
21
19
9
18
6
3
15
11
-26-
Partially Nested versus Fully Nested
Partially Nested (Curry):
e
Determine discrete allocations for each OD, based on expected
revenue from nested fare classes.
*Determine fare class booking limits within each OD allocation.
lili
UNwill.
-27-
Expected Revenue
per
O-D Pair BC
460
440
420
400
380
360
340
320
300
280
260
240
220
200
180
160
140
Seat
-28-
Fully Nested versus Partially Nested
3 Leg Example
Leg BC - Capacity=75
ODF
NDSP
Allocations
NDSP
BL
DOD-NFC
BL
ACY
ACM
ACB
ACQ
2
1
4
4
75
4
3
ADY
ADM
ADB
ADQ
3
0
0
0
BCY
BCM
BCB
BCQ
10
22
12
11
BDY
BDM
BDB
BDQ
6
0
0
0
38
31
15.
0
0
63
2
0
0
0
0
0
0
73
64
57
60
27
15
37
0
0
0
38
32
5
1
0
0
-29-
FULLY NESTED VS. PARTIALLY NESTED
2 Leg Flight
2.6
22
2
1.8
1.6
1.4
12
1
0.8
0.6
0.4
02
0
-02
0.7
0.74
0.78
0
0.82
0.86
Load Factor
DOD-NFC
+
NDSP
0.9
0.94
0.98
-30-
FULLY NESTED VS. PARTIALLY NESTED
3 Leg
Flight
3.5
3
C,,
:22
VZ.
E
oas
1.5
0
L.
.8.9
0
Load Factor
DOD-NFC
+
NDSp
-31
-
FULLY NESTED VS. PARTIALLY NESTED
4 Leg Flight
2.5
0
-0.5
-1
-
1.5
0.55
0.65
0.75
Load Factor
0
NDSp
+
DOD-NFC
0.85
0.95
-32-
Bid Price
e
Bid Price is a Shadow Price for the capacity constraints.
- Obtained from the same network formulations.
- The marginal value of the last seat of a given flight leg.
- Bid Prices establish a "cutoff" value for each flight leg,
on which decisions can be made whether to accept or
reject a given O-D/fare class request.
- For a single leg itinerary, a fare class is open for bookings
if the corresponding fare is greater than the bid
price, or shadow price, for the leg.
*For a multi-leg itinerary, fares must be greater than the sum
of the bid prices from the respective flight legs.
MI
-- No
-33-
3 Leg Example
Capacity= 7 5
B
A
A-B
B-C
C-D
C
34
197
169
BC:
197
AC
231
y
M
B
440
Y
M
B
Q
519
344
262
231
Q
315
223
197
D
AD:
400
582
379
302
269
-34-
DETERMINISTIC NETWORK METHODS
2 Leg Flight
2.6
2.4
22
Z2
1.8
1.6
1.4
12
1
0.8
0.6
0.4
0.7
0.74
0.78
0.82
0
~
0.86
Load Factor
BID
+
NDSP
0.9
0.94
0.98
m
-35-
SOLUTIONS
DETERMINISTIC 3 NETWORK
Leg Flight
32
3
2.8
2.6
2.4
22
3
2
-
1.8
0
1.6
1.4
C
12
-
0.8
C
u
0.6
0.4
02
a
0.76
0.84
0.8
0
0.88
Load Factor
BID
+
NDSP
0.92
0.96
-36-
DETERMINISTIC NETWORK METHODS
4 Leg Flight
2.6
2.4
22
2
1.8
1.6
1.4
12
1
0.8
0.6
0.4
0.55
0.75
0.65
0
Load Factor
+ BID
NDSP
0.85
0.95
-37-
Revenue Impa cts vs. Revisions
3 Leg Example
1
-
98%
Load
3
Number of Revisions
+ BID
NDSP
0
Factor
-38-
PROBABILISTIC NETWORK METHODS
2 Leg Flight
-1.5
-3.5
0.7
0.74
0.82
0.78
0
0.86
Load Factor
PBID
+
NPSP
0.9
0.94
0.98
ow
fifillh
-39-
PROBABILISTIC NETWORK SOLUTIONS
3 Leg
Flight
-0.5
-1.5
-
2.A
0.76
0.84
0.8
C
0.88
Load Factor
PSIC
+
NPSP
0.92
0.96
-40-
PROBABILISTIC NETWORK METHODS
4 Leg Flight
2
1
a
-1
0.55
0.65
0.75
0
Load Factor
PBID
+
NPSP
0.85
0.95
1,1111milmmillbi
-41-
UPPER BOUND
MEAN
36.12
9.94
18.61
34.06
ABY
ABB
ABM
ABQ
STD DEVI
14.91
13.92
16.01
-
25.96
ACTUAL DEMAND
ABY
ABB
ABM
ABQ
35
14
18
39
28
12
18
32
48
8
16
-42-
UPPER BOUND COMPARISON
2 Leg Flight
0.7
0.74
0.78
0
NDSP
0.82
0.86
Load Factor
0
+ BID
0.9
UPPER
0.94
0.98
__________N
-43-
UPPER BOUND COMPARISON
3 Leg Flight
0.76
0.8
0.84
0
NDSP
0.88
Load Factor
0
+ BID
0.92
UPPER
0.96
-44-
UPPER BOUND COMPARISON
4 Leg Flight
0.65
0.55
0
NDSP
0.815
0.75
Load Factor
0
+ BID
UPPER
0.95
U
-45-
Summary
* Nested Deterministic on Shadow Prices outperforms Nested
Probabilistic on Shadow Prices.
- Given full ODF forecasts, better to use a fully nested method,
such as NDSP, rather than a partially nested method.
e
Deterministic Bid Price approach performs well and uses a very
simple control methodology, however it is important to be
able to make frequent revisions using such an approach.
- Using Upper Bound, the true potential from better control of
seat inventories over current leg based approaches can be
determined.
Mhmssclwsetts
Institute of
Technology
Flight Transportation Laboratory-g
Planning and Scheduling of Tasks
ma
Dynamic Environment
Lyman R. Hazelton
23 May 1991
LRW9I-OO1
1VMassachusetts
Flight Transportation Laboratory-g
Institute of
Technology
The Strategic Control of systems requiring planning
and scheduling of activities is called
Operations Management
Reasoning about the future in a dynamic
environment.
Is
Determination of the time that a state or
process should be maintained.
I
Situation dependent objectives.
I
No final system State.
I
Often involve non-quantifiable parameters.
LRH-90-04"
Flight Transpotto
Institute of
~u~~i-~-
Adecision was made to attempt to solve the
problem with an "Expert Systems" approach.
However, existing Al planning methods
*
Were based on a back'-chained, goal seeking
technology.
Have been shown to be NP-hard or even
Non-terminating for. conjunctive goals.
Assumed a single actor, non-stochastic
universe.
*
Had no logic or even 'representation for
time, dependent activities.
In summary, the automatic reasoning technology
necessary to attack the problem did not exist.
LRH-90-043
Massachusetts
Institute of
T'-.chnology
Flight Transportation Laboratory--
At the time the research was initiated:
I
I
There were NO prograpis or even algorithms
for temporal database management
There were NO data representations for
concurrent temporally bo-anded, information
Automza± plan generatior. was restricted to
Single Actor
Determinate Dcmains
Instantaneous Actions
LRH4-90-050
Mesacwt
Ingtitute of
Flight Transportation Laboratory
Operations Management Model
LRII-90-{) 1'"
Technology
Flight Transportation Laboratory
Advice
Observations
A
Y
Temporal System Analyzer
RULESYS
(Logical Inference Engine)
Assertions
Schedules
Requests Y
ASchedules
Time Map Manager
(Temporal Database Manager)
TIMEBOX
SCHEDULE
LRH-9(~O('
Technogoy
Flight Transportation Laboratoryl
Pattern-I
Search Section
Pattern-3
Antecedent
Test Section
Rule
Consequent
Pattern-2
Test-1
Test-2
Assertion-1
Assertion-2
LRH-9
Technolo
Flight Transportation Laboratoryg
Plan: PAINT LADDER
Plan: PAINT CEILING
Procedure:
GET PAINT
GET LADDER
APPLY-PAINT LADDER
Procedure:
GET PAINT
Goal: NEAR CEILING
APPLY-PAINT CEIIJNG
Results
PAINTED LADDER
Rcsults
PAINTED CELMN
/
LRH-910-
Masschuset
isuen O
Flight Transportation. Laboratory
SHIT
Pan: PAINT CEILING
Han: PAINT LADDER
Procedur
GET PAINT
Goal: NEAR CEILING
APPLY-PAINT CEILING
Proceduim
GET PAINT
GET LADDER
APPLY-PAINT LADDER
Results:
PAINTED CEILING
Results:
PAINTED LADDER
LRHI-91-IY
Iitofd
Technology
Flght Transportation Laboratoryg
SKH
LIAction: GET PAINT
Action: GET PAINT
Action: GEf LADDER
Actiom GET LADDER
Action: CLIMB LADDER
Action: APY-PAINT LADDER
L Action: GEr PAINT
izzirizzi]
Action: GET LADDER
D
Action: APPLY-PAIN T CEILING
[Action: CLIMB LADDER
KIN
L
Action: APPLY-PAINT LADDER
_Action: APPLY-PAIN r CEIIuNG
Jom
LRH-91-0)i
Masschuisets
Institute of
Techftology
Flight Transportation Laboratoryg
Truth Maintenance
A first attempt to extend logic
into a dynamic environment.
,R:O
p
-
r 3R p q}T r
q
p
~q
R
p
T
V
r
r
q
LPKH-90-006
Institute of
Flight Transportation Laboratory
* Inferred evolution
R:p
-q
p
.+r-p-
q
{R p q} Tr
r (TM)
q
Rp
r p
Itieogy*gh
Transportation Laboratory~±
Introduce EXPLICITLY the
TIME INTERVAL during which
a proposition (was, is, will be)
true:
p (T)
where r is a time interval
having a starting time and
an ending time.
LRII-90-011
Massachugetts
Institute of
Technology
Flight Transportation Laboratory
Persistence:
R p
~p
5
q
q
(NO TM)
If it is raining (P), the roads
will be wet (q).
But if it stops raining CP),
the roads do not instantly
become dry. Wet roads persist.
LRH-90-010)
Masmichugetts
Institute of
-Techfiolo~v
Flight Transportation LaboratoryI.1
Temporal Logic (continued)
* Rules of Inference
Modus Ponens:
p
+q
)
p(
Non-persistent
q (start( ), o) Persistent
Sg (T)
0
0
0
LRI*-90-021
Massachusetts
Institute of
-Technology
Flight Transportation LaboratorylgInferred evolution -revisited
R:p(
1
) -q
r (Tl
T
T)-
,p(rfl T )
2
{R, p(T 1 ) q (T 2)} Tr (ri T2)
{R p(rTI) q (T2)} T p (Tifl T2)
~p (v)
qg(r) T
r(T,n T)
r(r)
p (T, n T
2)
LRI4-90-023
Mamschusetts
Institute of
atL
boatr
Flight Transportation
Laboratoryg
P
The problem stems from the
fact that the reasoner's
BELIEF (i.e., knowledge)
changes during the reasoning
process.
There are TWO tinie intervals
involved in temporal reasoning.
LKH-90-025
Massachusetts
Institute ofio
Flight Transportation Laboratorylg
* The ACTIVITY interval, during
which the proposition (was,
.is, will be) true
* The BELIEF interval, during
which the reasoner believes
a proposition about some
activity, interval to be true
LRH'-90-026
Institute
of
Tecbnolg
light Transportation Laboratory
Types of consequents
Bounded
Persistent_
Decayed
Probabilistic
LRH-90-042
-65I-.
I~'
BEFORE
12
ABUTS-BEFORE
- -- -il
I
P
(STARTS-BEFORE ENDS-DURING)
I'-.-]
(STARTS-BEFORE ENDS-EQUAL)
LYILJIJ
L3]
CONTAINS
F77T~1
(STARTS-EQUAL ENDS-DURING)
J7~
EQUAL
(STARTS-EQUAL ENDS-AFTER)
0
IS-CONTAINED-BY
I_
rx'zrn I
(STATS-DURING ENDS-EQUAL)
L
H
.J1Z17..
1,
*-
~
'-U--n
(STARTS-DURING ENDS-AFTER)
-1-
ABUTS-AFTER
[~j7fl
AMTER
Figure 3.2: Temporal Relations
6-8
'
.'
ZHj
Institute
of
Technology
Flight
Transportation Laborator
LRH1-904)41
Masschusetts
Institute of
Technlg
Flight Transportation Laboratory
FACT:
thing
attribute
Future
Past
t
t2
LRH-90-040
-68-
Most Recent
Belief
Earliest Belief
History
Belief Interval
- Owning Fact
Cell List
I
terval
U
PS4art Time
EImdA, Time
U
History Cell
Interval
S ftL
Owning History
Active Interval
Value List
Tin
m1
End Time
Value Cel
Value
Sense
Assumption
LRH-90-040
Massachusetts
Institute of
Technology
Flight Transportation Laboratory-g
CONTRIBUTIONS
Extension to Non-monotonic Temporal Logic
by introducing Belief Intervals
Introduced Persistence as rule specific knowledge
Designed structures to represent time dependent
knowledge
Implemented an efficient temporal database
management program
LRH-90-065
Massachusetts
Institute of
Technology
Flight Transportation Laboratory#
CONTRIBUTIONS
(continued)
e
Implemented a Temporal System Analyzer employing
Extended Temporal Logic and Persistence
Created a Scheduler Program, thereby extending
Domain Independent Planning to include
Parallel, Time Bounded, Non-Instantaneous Actions
LRH-90-06(
Massachusetts
ofi"'
T&e-hnoog
Flight Transportation Laboratory-
g'
Novel ideas and methods developed for this system
include
*0 A highly compact representation for the
description of descrete time dependent
processes.
*
An efficient time based logical inferrence
system.
Deeper understanding of human cognitive
and communication processes involved in
Command and Control Systems.
*
A replacement of "Truth Maintenance" by
"History Maintenance", and a better
understanding of default versus dynamic
logic.
LRH-90-044
Massachusetts
Institute of
TectInology
Flight Transportation Laboratory'
0
LRH-90-068
Concentration in U.S. Air Transportation:
An Analysis of Origin-Destination Markets
since Deregulation
Jan Van Acker
Flight Transportation Laboratory
May 23, 1991
Agenda
I. Thesis Objective and Methodology
II. Analysis of Top 100 Markets
III. Analysis of Dominated City Markets
IV. Conclusions
I. Thesis Objective
e
Study effects of deregulation on concentration
- Focus on Origin-Destination City-Pair Markets
Focus on Concentration in O-D City-Pairs
e
e
Other studies found:
-
Fares are positively related to concentration
-
Concentration levels have decreased on average
Our study looked at:
- Top 100 domestic O-D markets
- Markets out of dominated cities
Measurement of Concentration
e
Concentration indices used:
- Hirschman-Herfindahl Index (HHI)
-
-
e
2-Firm Concentration Ratio (C2)
Number of Competitors with >5% Market Share
(Number of Effective Competitors)
Market share is measured in terms of local
passengers transported in market
II. Changes in Concentration in Top 100
Markets
e
Markets ranked 1-100 in terms of local passengers
transported in 1989
e
Cumulative number of passengers was 31 % of U.S.
domestic total in 1989
e
Years studied: 1979, 1981, 1983, 1985, 1987, 1989
e
With focus on 1979, 1985, 1989
Average Number of Effective Competitors was
One more in 1989 than in 1979
Year
Average Number of
Effective Competitors
1979
1981
1983
1985
1987
1989
2.7
3.3
3.5
3.8
3.6
3.7
4.51-
2.5EF
1.5 F
1979
1981
1985
1983
Year
1987
1989
56 Markets Were Served by Four or More Effective
Competitors in '89, as Compared to only 16 in '79
# Carriers
With >5% MS
1979
1985
1989
1
2
3
4
8
38
38
11
5to6
5
7to8
0
0
17
30
24
28
1
1
19
24
29
24
3
501r
40
30
gd
10
0
1979+ 1985*
1989
# of Carriers with >5% Market Share
62% of the Passengers Flew in Markets Served by 4 or
More Effective Competitors in '89
- -
only 18% in '79
# Carriers
With >5% MS
1979
1985
1989
1
2
3
4
5 to 6
7 to 8
5.5%
30.9%
45.8%
11.3%
6.5%
0.0%
0.0%
12.6%
29.1%
26.3%
30.7%
1.2%
0.7%
16.5%
20.5%
28.9%
30.2%
3.3%
40%
30%
10%
0
6
4
2
a 1979+ 1985* 1989
# of Carriers with >5% Market Share
8
10
Average HHI Was Lower in 1989 than in 1979
Year
Average HIHI
1979
1981
1983
1985
1987
1989
4917
4077
3913
3361
3705
3586
5.515F
U
'U
4.5 F-
0
3.5 F
2.5 1979
1981
1983
Year
1985
1987
1989
The Majority of the Markets Experienced a Decrease
in HHI from 1979 to 1989
Change in HHI
-8000 to -6000
-6000 to -4000
-4000 to -2000
-2000 to O
0 to 2000
2000 to 4000
4000 to 6000
Total Decreased
Total Increased
Average Change
-7
-?
o
-5
-3 3
T-1
1
3
(Thousands)
1979-1989 1979-1986 1985-1989
Change in HHI
The Non-Hub Markets Were Served on Average by a
Greater Number of Effective Competitors in '89 than
the Hub Markets
Year
Hub
Markets
Non-Hub
Markets
1979
1981
1983
1985
1987
1989
2.7
3.3
3.7
3.9
3.5
3.5
2.6
3.4
3.4
3.7
3.7
4.0
4.5
3.51F
31-
1.5 F
I
I
1979
1981
1983
1985
1987
0 Hub Markets Non-Hub Markets +
1989
Concentration Decreased from '79 to '89 in All but
One of the Non-Hub Markets
'85-'89
'79-'89
'79-'85
-4to-3
-2to-1
0
1 to 2
3to4
5to6
0
1
8
33
7
0
0
5
11
24
9
0
0
11
19
17
2
0
Total Decreased
Total Increased
1
40
5
33
11
19
Average Change
1.10
0.33
1.43
Change in # Carriers
With >5% MS
But Was Higher in '89 than in '79 in 30% of the Hub
Markets
Change in # Carriers
With >5% MS
'79-'89
'79-'85
'85-'89
1
11
9
4
1
0
7
8
28
8
0
1
26
16
8
0
0
Total Decreased
Total Increased
12
30
7
36
27
8
Average Change
1.12
-0.41
0.71
-4to-3
-2 to -1
0
1 to 2
3to4
5to6
25
The Top 10 Markets Were on Average Less
Concentrated than the Top 50 and Top 100 Markets
Year
Top 100
Markets
Top 10
Markets
Top 50
Markets
1979
1981
1983
1985
1987
1989
86.6%
79.2%
78.4%
73.3%
74.7%
73.6%
79.5%
74.8%
75.0%
70.6%
72.2%
66.2%
83.9%
77.7%
77.0%
71.8%
74.3%
73.3%
100%
95%90%85%80%75%70%65%60%
1979
1981
1983
1985
1987
o Top 100+Top 10*Top 50
Year
1989
Conclusions of Top 100 Markets Analysis
e Average concentration was lower in '89 than in '79
e
Concentration was lower in 70% of the markets
e
Non-hub markets were better off on average in 1989
than hub markets
e
Top ten markets were less concentrated on average
than top 100 markets
III. Changes in Concentration in Top Ten
Markets out of Dominated Cities
e
Cities at which 60% of total passenger enplanements
in 1985 were carried by one airline, or 85% by two:
Atlanta
Charlotte
Cincinnati
Dayton
Denver
e
Detroit
Greensboro
Memphis
Minneapolis
Nashville
Pittsburgh
Raleigh/Durham
St. Louis
Salt Lake City
Syracuse
Markets ranked 1-10 in terms of local passengers
transported in 1989 out of each of the cities
Changes in Concentration in Top Ten
Markets out of Dominated Cities
e
Years studied: 1979, 1981, 1983, 1985, 1987, 1989
e
With focus on 1979, 1985, 1989
Average Number of Effective Competitors in 150
Markets Peaked in '85, but Was Still Higher in '89
than in '79
Year
Dominated
Airport Markets
1979
1981
1983
1985
1987
1989
2.2
2.8
2.8
3.1
2.9
2.5
4.5
4
3.51F
3F
2.51F
0",
1979
1981
1983
Year
1985
1987
1989
Average Number of Effective Competitors for each
of the Dominated Cities
Top 10 Atlanta Markets
Top 10 Charlotte Markets
3.
4.5
4.0-
4.0-
3.5-
3.5-
3.0
3.0
.............
...........
.- - v
2.52.0-
2.52.01.5-
1.0'
5.0
1979
'op
1981
1983 1985
Year
1987
1989
1979
10 Cincinnati Markets
1981
1983 1985
* Year
1987
1989
Top 10 Dayton Markets
4.5-
4.5-
4.0
4.0-
3.5 -
3.5
3.0
3.0-
2.5
2.5-
e--"...............................
2.0 t2.0-
1979
1981
1983 1985
Year
1987
1989
1979
1981
1983 1985
Year
1987
1989
Average Number of Effective Competitors for each
of the Dominated Cities
Top 10 Denver Markets
Top 10 Detroit Markets
4.5
4.0-
4.
3.5-
3.
3.0
3.0
2.5
2.52.0-
2.05
1979
1981
1983
1985
Year
1987
1989
0
1.4
1979
Top 10 Greensboro Markets
5.Oi-
1981
1983 1985
Year
1987
1989
Top 10 Memphis Markets
Or4. 5
4.0
3.0
3.5
[
0
2.5-
.
.......
..
...
..........
2.5
2.01.
1979
1981
1983 1985
Year
1987
19899
1979
1981
1983 1985
Year
1987
1989
Average Number of Effective Competitors for each
of the Dominated Cities
Top 10 Minneapolis Markets
Top 10 Nashville Markets
5.0
4.51
4.0 1
3.01.
2.546/
1979
1981
1983
1985
1987
1989
1979
1981
Year
Top 10 Pittsburgh Markets
5.%#1
To~
To
4.5
4.5 -
4.0
4.0-
3.5
33.5-
3.0
3.0-
2.5
n
2
10
1983
Year
1985
1987
1989
Raleigh-Durham Markets
2.5-
__z
2.0
2.0-
1.5
1979
1981
1983
1985
Year
1987
1989
1979
1981
1983
Year
1985
1987
1989
Average Number of Effective Competitors for each
of the Dominated Cities
Top 10 Salt Lake City Markets
Top 10 St. Louis Markets
5.0
4.5-
4.5
4.0
4.0-
3.5-
3.5 -
3.0-
3.01-
2.5-
2.5
2.0-
2.0
1.5
1.5
1n
-
.
1979
1981
1983 1985
Year
1987
1989
1979
1981
1983 1985
Year
Top 10 Syracuse Markets
1979
1981
1983
1985
Year
1987
1989
1987
1989
Changes in Concentration in the Top Ten Atlanta
Markets
O-D City-Pair Markets
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Boston
Chicago
Dallas/Fort Worth
Los Angeles
Miami
New York
Orlando
Philadelphia
Tampa
Washington
HHI
1989
'79-'89
'79-'85
'85-'89
5446
2949
5932
5089
3737
3913
5608
4097
6002
4701
455
-2538
-350
-219
-1004
-935
482
-995
955
-193
-547
-2031
-1968
-656
-885
-1294
-880
-410
-637
-721
1002
-508
1618
437
-119
359
1362
-585
1593
528
7
3
10
0
3
7
-434
-1003
569
Total Decreased
Total Increased
Average
Change in HHI
4747
Concentration Levels Decreased Substantially in
Most of the Top Ten Syracuse Markets
O-D City-Pair Markets
Syracuse
Syracuse
Syracuse
Syracuse
Syracuse
Syracuse
Syracuse
Syracuse
Syracuse
Syracuse
Atlanta
Boston
Chicago
Detroit
Los Angeles
New York
Orlando
Philadelphia
Tampa
Washington
HHI
1989
'79-'89
'79-'85
'85-'89
4418
9045
4119
8942
1585
5820
3047
9741
2695
8289
-5552
175
-5473
3967
-5234
653
-6014
790
-5875
-1406
-4065
-4091
-5889
-740
-4924
-1756
-5507
493
-5178
-4879
-1487
4266
417
4707
-310
2409
-507
296
-697
3473
6
4
9
1
4
6
-2397
-3653
1257
Total Decreased
Total Increased
Average
Change in HHI
5770
Concentration Increased in all Top Ten St. Louis
Markets after the TWA-Ozark Merger
O-D City-Pair Markets
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
Chicago
Dallas/Fort Worth
Denver
Detroit
Houston
Los Angeles
New York
Phoenix
San Francisco
Washington
HHI
1989
'79-'89
'79-'85
'85-'89
3347
4528
4671
3306
3561
5486
8860
4780
6567
8252
-2161
-506
-238
-2180
-916
-48
2650
-320
155
302
-3019
-2037
-2333
-2200
-1626
-654
-1356
-1360
-547
-3798
858
1530
2095
20
710
606
4006
1040
701
4100
7
3
10
0
10
-326
-1893
1567
Total Decreased
Total Increased
Average
Change in HHI
5336
Conclusions of Dominated City Markets
Analysis
e
Single trend of hub development led to decreased
concentration through '85 at most of the cities,
but to increases from '85 on
e
Two-hub markets were less concentrated than
one-hub markets in 1989
* Average concentration across the 150 markets was
slightly lower in '89 than in '79
IV. Conclusions
e
Concentration was lower in top 100 markets, both
on average and in most of the markets
e
Concentration in non-hub markets decreased
throughout period '79-'89 because of
development of hub-and-spoke networks
" These networks led to increases in concentration in
most hub markets after 1985
CD
Conclusions
e
Single trend of hub development led to decreases in
concentration through '85 at most of the
dominated cities, but to increases from '85 on
e
Concentration was on average slightly lower in the
150 markets out of dominated cities in 1979 than
in 1989, and was lower in half of the markets
-102-
Pricing in the Airline Industry
Current Practice and
Future Research
Theodore C. Botimer
MIT Flight Transportation Laboratory
Presentation to Cooperative Research Program
May 23, 1991
Ihi,
I1,
1111
10111NIM1111
kilmhIIIW1,W
-103-
Presentation Outline
Overview
* Nature of Airline Competition
* Fare Product Differentiation
- Seat Inventor y Management
* Pricing Strategies
- Role of the Pricing Analyst
- "The Ultimate Pricing Model"
* Theoretical Issues for Investigation
Case Study Analysis
* Case Study Overview
* Case Study Objectives
* O/D Market Choice
- ATL - BOS Market
- ATL - STL Market
- Conclusions
-104-
Nature of Airline Competition
-
Hub and spoke route structures prevail in
the industry allowing almost every major
carrier to serve any O/D market
-
Most competition on non-price level
- Dollar value of nonstop service is unclear
-
Must consider strength of competitive
position in e ach O/D market separately
-
Characterize c ompetitic n in all markets:
- major players
level of service offered
- number of flights p r day offered
- nonstop vs. nonsto p competition
e
-
Anticipate r espor se to price changes:
* who ai oe the competitors?
* do the competitors offer comparable
service in the market?
- how have competitors reacted to
fare chan ges in the past?
- what response will be given to hostile
reactions by competitors?
MOMME1111"
,1,1,,
illdilill
III,
-105-
Fare Product Differentiation
-
Airlines seek to segment demand by
off ering differentiated fare products in
different fare classes
-
Delta offers tickets in 10 fare classes:
1) F - full fare first class
2) Y - full fare coach class
3) B - reserved for military/
convention/negotiated fares
4) M - highest discount coach fare
5) H - discount coach class fare
6) Q - discount coach class fare
7) K - reserved for competitive filings
8) L - reserved for competitive filings
9) A - first class free tickets
10) W - coach class free tickets
-
Differentiation occurs within fare classes
* peak vs. off-peak fares
- weekday vs. weekend fares
-106-
Fare Product Differentiation (con't)
-
Fare restrictions or "fences" used to
control which type of consun er is able to
purchase which type of ticket
-
fare restriction s include:
advanced purchase requirements
Saturday night stayover
blackout periods
flight validity resi:rictions
(good for travel between...)
ticket purchase r estrictions
(purchase ticket s by...)
availability limits for discount fares
military discount fares
senior citizen discount fares
Common
-
.....
WANNME,
-107-
Seat Inventory Management (IM)
- Pricing sets O/D prices and restrictions
-
IM decisions made with fixed prices and
restrictions
-
IM seeks to maximize revenue given
fixed prices and restrictions
-
IM controls price/seat quantity decisions
- protect full far'e seats
- limit discount fares
- strictly limit deep discount fare
seat availability
-
Matching stances require booking limits
- strictly limited availability on
competitive far e filings
-108-
Pricing Strategies and Their Effects
- Matching a fare
-
retain market share
possible drop in yield
remain listed on Page 1 of CRS
often done to remain competitive
viewed as price taker in the market
- Not matching a fare
-
possible loss of market share
maintain yield
may lose competitiveness
loss of goodwill
- Partially matching a fare
e
-
attempt to retain market share
reduce non-matching yield loss
market factors influence strategy
will be non-competitive at peak
accept that competitor offers low
fare on all flights
WWSMN WIININ
1114
, ,, IIN
i,
-109-
Role of the Pricing Analyst
-
Analysts do not look at operating costs
-
Consider strength of competitive position
in each O/D market
-
Add routing restrictions to discount fares
-
Pricing analysts should be familiar with
own market and relevant hub:
- traffic flows
- flight load factors
-
Be aware of fare differential effects
- high differentials not seen on CRS
- business travelers susceptible to
higher differentials
- not all fares registered in ATP
listings are available in reality
-
Must monitor the number of bookings to
determine the effect on yield of changes
-110-
Ultimate Pricing Model
-
Inputs:
- published daily fare changes
- system-wid[e flight schedule
- price level (by O/D market & flight)
-
Outputs:
-
Su "
-
ested
mach
gb
strategy
~
in
g
--
partial matching
--
not matching
- Projected impact on
- Projected impact on
- Management reports
-suggested match
market share
revenue
telling:
ing decision
-- implemented matching decision
for matching decision
- Ability to run simulations
- Ability to do what-if scenarios
-- reasons
NwA
1011ME111
1,
-111-
Theoretical Issues for Investigation
-
Joint seat/price optimization problem
-
Optimal differential pricing strategies
-
Model development for pricing strategies
- matching
- not matching
- other pricing strategies
of price changes
-
Impacts
-
Measurements of price elasticity
-
Explore impacts of pricing strategies on:
-
profitability
load factor
yield
customer satisfaction
-112-
Case Study Overview
-
Close look at 10 O/D markets
-
Representative cross section of markets
flown by Delta Airlines
-
Quarterly analysis
-
Examine quarterly data 1986:1
-
Give consideration to:
0
blI
p
4
s
eF6L&
-
1990:2
aresa
- competitive responses
- major price level changes
-
Use information from several data bases:
- PIPPS (Historical ATP data)
- DOT O/D traffic stats (10% sample)
- Official Airline Guide
-
Preliminary analysis on two markets:
- ATL - BOS
- ATL - STL
, I,
IM111
, , ''111,111MI
IN
I,
-113-
Case Study Objectives
-
Initial look at revenue management from
pricing perspective
-
Develop market by market case studies
- Present a market overv iew
- Characterize pricing practices
- Analyze competitive environment
- Uncover competitive characteristics
- Highlight major market events
the quality and level of detail of
the ava ilable data sources
-
Analyze
-
Relate
-
Develop
market strength to fare level
- between carriers
- over time
a measure of the sensitivity of
travelers to changes in fare level
- Determine selling fares during the period
-
Use available data to determine the
effects of pricing decisions
- Discuss future directions for research
-114-
O/D Market Choice
Lenath of Haul
-
Short Haul (<1000 miles)
Medium F aul (1000-2000 miles)
-
Long Haul (>2000 miles)
-
Nature of Competition
of Competition
-
Delta offers non-stop service
Only competitors offer non-stop service
- No one offers non-stop service
Markets Chosen
1)
2)
3)
4)
5)
6)
7)
8)
9)
10)
ATL-BOS
ATL-SEA
ATL-STL
BOS-PHX
CLT-MSP
DFW-PHL
JAN-SDF
MSP-SAN
MSY-PWM
SAV-SAN
-115-
ATL - BOS Market Characteristics
1986:1 - 1990:2
-
Two non-stop carriers during the period
--
Delta
--
Eastern
-
Non-stop carriers flew 93% of all pax
-
Frequency of approximately 12 daily
non-stops each way
-
Total traffic leve l of 925 1passengers
per day in both directions
-
Carriers with ATL hub
--
Delta
--
Eastern
- Eastern Airlines strike in 1989:2
-116-
ATL-BOS Passengers
1986:1 - 1990:2
11
10 -
9
I
8
8
8
7:3I
I
I
I
I
I
I
I
I
I89:
90:
I
89:1 89:3 90:1
88:3
86:3 87:1 87:3 88:1
86:1
90:2
88:2 88:4 89:2 89:4
87:2 87:4
86:2 86:4
YearQuarter
,,
111h
B11141111111
kllilfillllll
-117-
ATL-BOS Average Fare
1986:1 - 1990:2
220
210
200
190
180
170
160
150 1140
I
130
86:1
863
86:2
87:1
86:4
87:3
87:2
88:1
87:4
88:3
88:2
Year.Quarter
89:1
88:4
I
89:3
89:2
t
I
90:1
89:4
90:2
MVLBOS Morke~
/1A
Shore
8OL&
60
If)
nfl
24
20v
10
-A
86:1
~
86:2
86:3
I
I
86:4
87:1
--
87:2
87:3
87:4 88:1
882
88:3 88:4 89:1
YearQuarter
-,-
Delta
-
-
Eastem
89:2
89:3 89:4 90:1
902
/YL~HS
verae rare
1986:1
260
-~----
m-
-
-
1990:2
_
--
-
--
_-
-
-
_
_
240
20
140
120
86:1
86:2 86-3 86:4 87:1
87:2
Delta
87:3 87:4 88:1
88-2
88:3
YearQuarter
Eastem
M84
89:1
89:2 89:
894 9- 19:
ATL - BOS
Summary Table
Passengers
Delta 88:2
Delta 89:2
% Change
Eastern 88:2
Eastern 89:2
% Change
OA 88:2
OA 89:2
% Change
Market 88:2
Market 89:2
% Change
6118
7676
25.47
3508
9
-99.74
615
1387
125.53
10241
9072
-11.41
Revenues
1152456
1584791
37.51
599469
1674
-99.72
88075
214329
143.35
1840000
1800794
-2.13
Average
Coupon
Passengers
Market
Revenue
Yield
Fare
Mileage
Per Day
Share
Share
Per CPM
188
206
9.60
171
186
8.84
143
155
7.90
180
199
10.48
5872057
7372559
25.55
3348572
8514
-99.75
653125
1429440
118.86
9873754
8810513
-10.77
67.2
84.4
25.47
38.5
0.1
-99.74
6.8
15.2
125.53
112.5
99.7
-11.41
59.74
74.95
25.47
34.25
0.09
-99.74
6.01
13.54
125.53
100.00
100.00
0.00
62.63
88.01
40.51
32.58
0.09
-99.71
4.79
11.90
148.65
100.00
100.00
0.00
19.63
21.50
9.53
17.90
19.66
9.83
13.49
14.99
11.19
18.64
20.44
9.68
W
ATL - STL Market Characteristics
1986:1 - 1990:2
-
Four
non-stop carriers during the period
--
Delta
--
Eastern
--
Ozark
-- TWA
-
Non-stop carriers flew over 90% of all pax
-
Frequency o f approximately 15 daily
non-stops ea ch way
-
Total traffic level of under 450 passengers
per day in both directions
-
-
Carriers with ATL hub
--
Delta
--
Eastern
Carriers with STL hub
--
Ozark
-- TWA
-
Eastern Airlines strike in 1989:2
-
Ozark - TWA merger in 1987
-122-
ATL
-
STL Passengers
1986:1
-
1990:2
5000
4500
8 4000
3500
30
00
'
'
f
I
I
I-
-1'
l
I
I
I
I
86:1
86:3
87:1
87:3
88:1 88:3
89:1
89:3
90:1
86:2
86:4
87:2 87:4
88:2
88:4
89:2
89:4
90:2
Year.Quarter
.. Ww
m
IN
-123-
ATL
-
STL Average Fare
1986:1
-
1990:2
180
170 160 150 140 130 120 110 100'
'
'
'
90:1
89:3
89:1
88:3
88:1
87:3
87:1
86:3
86:1
90:2
89:4
89:2
88:4
88:2
87:4
87:2
86:4
86:2
YearQuarter
ATL - SIL- Market Share
1986:1
-
1990:2
50r
Q-
L-)
30
\
20 :x
10
~
86:1
86:2
88:3
86:4 87:1
87:2
DL
87:3 87:4
88:1
88:2
88:3 88:4
Year:Quarter
2 1# -0Z
89:1
89:2 89:3
89:4 90:1
90:2
ATL - STL Average Fare
1986:1 -1990:2
IOU
170 r
160 -
CDe
[~'
4>
en
14O
130
120
110
100
90
L
..
86:1
.
i
86:2 86:3 86:4
--. I
87:1
-
I
87:2
-
-L
87:3
87:4 88:1
._
-
882
_I-
88:3
--
-1
..
L0
lw
...
88:4 89:1
Year.Quarter
Di
El-
-.
0-1
..
..
-
.
-
-
-L
89:2 89:3 89-4 90:1
_-
J
-
90:2
ATL - STL
Summary Table
Passengers
Delta 88:2
Delta 89:2
% Change
Eastern 88:2
Eastern 89:2
% Change
TWA 88:2
TWA 89:2
% Change
OA 88:2
OA 89:2
% Change
Market 88:2
Market 89:2
% Change
1799
1882
4.61
1049
1
-99.90
1155.00
1631.00
41.21
67
57
-14.93
4070
3571
-12.26
Revenues
259057
320589
23.75
149500
94
-99.94
172264.00
261282.00
51.68
6003
6560
9.28
586824
588525
0.29
Average
Coupon
Passengers
Market
Revenue
Yield
Fare
Mileage
Per Day
Share
Share
Per CPM
144
170
18.29
143
94
-34.04
149
160
7.41
90
115
28.45
144
165
14.30
881790
919694
4.30
510126
484
-99.91
560872.00
793400.00
41.46
48980
38358
-21.69
2001768
1751936
-12.48
19.8
20.7
4.61
11.5
0.0
-99.90
12.7
17.9
41.21
0.7
0.6
-14.93
44.7
39.2
-12.26
44.20
46.24
4.61
25.77
0.02
-99.90
28.38
40.07
41.21
1.65
1.40
-14.93
100.00
100.00
0.00
44.15
54.47
23.39
25.48
0.02
-99.94
29.36
44.52
51.68
1.02
1.11
8.96
100.00
100.00
0.00
29.38
34.86
18.65
29.31
19.42
-33.73
30.71
32.93
7.22
12.26
17.10
39.54
29.32
33.59
14.59
W.W
w
iili,
-127-
Conclusions
-
Carrier strength varies by O/D market
-
Delta holds a stronger position in ATL - BOS
than in ATL - STL
-
Delta fare levels may have been too high
in ATL - STL during the strike given its
competitive position
-128-
Future Directions
-
Quantify consumer price sensitivity and
market share chan ges
-
Determine relationships between
market
strength arnd fare levels
-
Develop a model to characterize competitive
structure of markets
-129-
CHANGES IN O-D PASSENGER TRAFFIC FLOWS
NEWARK AIRPORT
Chung Y. Mak
and
Professor Peter P. Belobaba
MIT Flight Transportation Laboratory_.
MIT / Industry Cooperative Research Program
Annual Meeting
May 24,1991
-130-
BACKGROUND : PREVIOUS ANALYSIS
*
Removal of PeoplExpress from the New York (EWR)
market has had the most significant impact on traffic
flows.
*
Domestic connecting passengers have dropped in both
absolute and percentage terms at all three airports,
suggesting a shift by carriers away from New York
airports as domestic hubs.
Newark Airport (EWR)
*
Stable departure levels since PE withdrawal, but fewer
seats and reduced aircraft sizes.
*
Major drop in on-board passengers after 1986-3;
downward trend continues through 1989-3 for virtually
all carriers.
*
Local originating passengers cut by half when PE failed;
levels have barely returned to pre-1984 levels.
-
Domestic connecting passengers were similarly affected
by PE withdrawal from EWR.
.-
u.y
.
....
Ili
Ii
-131-
Total Seats Departed for Majors
Al New York City Areo Akports
12.21211.811.611.411.211 10.810.610.410.2109.8 1
9.6-1
1084 2
1
3
1
1
1
4 1085 2
1
3
4 1086 2
3
Ouarter
Figure 1.2
4 1087 2
3
4 1088 2
3
4 1089
-132-
Total Seats Departed for Majors (T9)
Newcrk 1I*wnflon
6
5.8
5.65.45.25 4.84.64.44.24
3.6
1
2
3
1984
4
1
2
3
1985
4
1
2
3
1986
Ouartr
Figure 1.8
4
1
2
3
1987
4
1
2
3
1988
4
1
-133-
Total Seats Departing
Nwark Interationa
3.2-
3 2.82.6 2.4 2.221.8
S=*
1.6
1.4 1.2 0.80.60.40.20
1
2 3
1984
4
1
2 3
1985
4
1
2 3
1986
Quarter
O
co
Figure 1.9
4
+
1
PE
2 3
1987
4
1
2 3
1988
41
-1-1--
"imbMWA&
-134-
Total Onboard Pax for Majors
Newark International
43.93.8 3.73.63.53.4-
3.3 X
ac
M=
03.
3.23.1 -
33-
2.92.82.7 2.2.5 2.42.3
1
2
3
1984
4
1
2
3
1985
4
1
2
1986
Quarter
3
Figure 2.3
4
1
2
3
1987
4
1
23
1988
4
1
-135-
Pcx
Total Onboard
Nq..ar
Intewngtnaw
2.1
1.9
25
1.4
-
1.7 -
A'
\
//
./
/
/
x
L ~
0.5
-
0.7
-
0.6
0.7
~
0.3
0.3 r.3
2
4.
2
!
-
Figure 2.6
.6i 3 i
4.P
z~ I4;~i
2
4.
.3
4
Total Cninecting Pax (EWR)
(Ten Percent Sample)
10 -
81-
1Q84
3
1Q85
3
186
3
Quarter
Figure 3.5
1Q87
3
1Q8
3
1Q89
, ,,
IIIIIMI
IJwmIIwIiIm
-137-
Total Connecting Pax (EWR)
5
(Ten Percent Saample)
4.5
4
3.5
1=
00
1.5
1
0.5
0
Quarter
0
CO
+
Figure 3.8
PE
MImIMi
,
-138-
NEWARK AIRPORT
TRAFFIC FLOW ANALYSIS (Phase 2)
ANALYSIS ( hase 2)
OBJECTIVE
*
Identify and evaluate changes in O-D passenger traffic
flow patterns through Newark (EWR) and alternative
hub routings.
Determine shifts in connecting traffic away from EWR in
O-D markets previously served by PeoplExpress.
HISTORICAL DATA
*
Ten percent ticket coupon sample provides passenger
itinerary information by quarter from 1985 to 1989.
e
Database Products Inc. "OD Plus" database used to
extract data.
-
Official Airline Guide (OAG), schedule data for each of
the periods.
INN
IIII
-139-
SSENGER
R TRAFFIC FLOW ANALYSIS
PA
P ARR F N G F TRAFFIC FLOW
DEMAND AND SUPPLY MEASURES
-
Ten percent O-D passengers travelled between each
selected city pair by carrier.
.
Scheduled service in each city pair by carrier.
AIR CARRIERS
-
"Major" U.S. carriers offering service to domestic
destinations, defined to include smaller airlines with
large market presences (e.g. Midway).
-140-
ANALYSIS METHODOLOGY
*
Obtained top 500 US Domestic O-D markets in terms of
passenger traffic for 1989.
*
Selected markets served by PeoplExpress in 1986.
-
Discarded all city pairs with New York as an
Origin/Destination, leaving 50 sample markets.
*
Used O-D Plus to obtain passenger traffic data for 3rd
quarter 1986 for all major carriers serving these city
pairs.
*
Selected O-D pairs based on market share and passenger
information for detailed analysis:
-
markets with greater than 5% market share
by PeoplExpress in 1986 or;
-
markets with more than 20 passengers
carried by PeoplExpress per day.
*
A total of 20 markets were chosen based on these criteria.
*
Used O-D Plus again to obtain detailed passenger traffic
information by individual market and carrier from 1985-3
to 1989-3.
-141-
20 SELECTED O-D MARKETS
CHI-BDL
ORL-CMH
PIT-HOU
PIT-LAX
WAS-MIA
WAS-BUF
WAS-DEN
WAS-PVD
BOS-CHI
BOS-DFW
BOS-DET
BOS-FMY
BOS-HOU
BOS-LAX
BOS-ORL
BOS-PIT
BOS-SFO
BOS-WAS
BOS-DEN
BWI-DEN
1986-3
PE Share
1986-3
3.30%
2.25%
4.95%
4.91%
5.74%
9.36%
8.73%
4.99%
6.02%
4.18%
2.36%
5.16%
5.41%
1.14%
1.80%
16.45%
1.86%
3.44%
12.10%
6.45%
5440
280
3240
1810
7950
3500
7500
1670
11960
2960
2120
570
2600
1920
2110
9490
2530
9220
9320
1930
Pax/Quarter
-142-
FINDINGS
Aggregate : 20 O-D Markets
*
Total traffic in selected O-D pairs decreased slightly
since withdrawal of PeoplExpress in 1986-3.
-
aggregate traffic decreased by 5.94% from 1986-3 to
1989-3.
*
However, proportion of this traffic connecting through
EWR dropped from 4.84% to 0.71% during the same
period.
-
In 1985, PeoplExpress carried 8% of total traffic in these
markets.
-
By 1989, Continental carried a total of 10% of traffic in
these markets.
e
However, only 1% was carried by CO via EWR.
,.iiIIMEMII Nl
DIAIN.,
-143-
Total Passengers for 20 Selected Market
10% Sample
200
150
c
100
50-
0
1985
1987
Year
1986
_-_ Total
* EWR
1988
1989
-144-
Total Market Share
20 Selected Markets
15%
10%
0%
1985
-- _ CO - EWR
O
1986
1987
Year
CO - OTHER
a
PE
A
1988
E3
1989
TOTAL - EWR
-145-
Disaggregate Market Analvsis
Examples of market share changes 1985 to 1989 follow,
showing PE, CO and the two competing carriers with the greatest
increase in market share:
-
"CO - Other" refers to Continental traffic routed
primarily through other CO hubs.
-146-
Market Share Comparison
BOS - PIT
100.00%
90.00%
80.00%
70.00%
60.00%
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
1985
1986
_ CO - EWR
C_ PA
1987
Year
* CO - OTHER
1988
A
PE
+ US
BOS-PIT:
PE had 28% market share in 1985, virtually all of which
was taken over by USAir (non-stop service).
-
CO never recaptured significant market share.
1989
1161
,1,, 11,
,11,11M
INI
-147-
Market Share Comparison
BOS - FMY
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
CO - OTHER
CO - EWR
o
.*
AA
1988
1987
Year
1986
1985
I PE
DL
BOS-FMY :
PE had peak market share of 45% in 1985, of which CO
now carries only 9% via EWR.
-
AA market share grew from 0 to 24% (CNX via RDU).
.
DL also took over market share (via ATL and CVG).
1989
-148-
Market Share Comparison
BOS - DEN
50.00%
...
..
....
-
40.00%
30.00%
20.00%
-
10.00%
>0
0.00%
1985
1986
1988
1987
1989
Year
_
CO - EWR
-3 ML
CO - OTHER
+
I
PE
UA
BOS-DEN:
CO has captured most of PE's 12% market share, but on
non-stop service. UA also shows market share growth
(non-stop service).
mms
=Ilml
Ni
-149-
Market Share Comparison
BOS - WAS
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
CO - OTHER
CO - EWR
7,T A
1988
1987
Year
1986
1985
--
;j PE
US
BOS-WAS:
-
PE had 10% market share in 1985. CO did not capture
any of this traffic (via EWR), except in 1987 when CO
offered non-stop service to IAD.
Greatest MS growth by US and UA (both non-stop
services).
1989
-150-
Market Share Comparison
CHI - BDL
80.00%
70.00%
60.00%
..
...........
50.00%
dFW
40.00%
30.00%
20.00%
10.00%
0.00%
1985
1986
1987
1988
Year
CO - EWR
__NW
CO - OTHER
& PE
g UA
CHI-BDL:
PE's 9% market share in 1985 was captured by CO via
EWR until 1988, when UA increased non-stop service.
1989
WIMN
111M
'I'Allul
L
-151-
Market Share Comparison
PIT -LAX
80.00%
70.00%
60.00%
50.00%
40.00%
-
30.00%
20.00%
10.00%
2
-
0.00%
_
CO - EWR
1988
1987
Year
1986
1985
-
CO - OTHER
& PE
1_ ML
PIT-LAX.
.
PE carried up to 5% of market share in 1986 via EWR.
-
CO increased its MS from 0 to 16% in 1987, but not via
EWR (i.e. via IAH, DEN, CLE).
1989
-152-
Market Share Comparison
WAS - BUF
90.00%
80.00%
70.00%
60.00%
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
1985
1986
1987
1988
Year
CO - OTHER
-- CO - EWR
_3 UA
--
2 PE
US
WAS-BUF:
-
PE carried 37% of market share in 1985, only 7% of which
was captured by CO via EWR in 1987.
-
Biggest market share gains went to UA (non-stop to IAD)
and US (non-stop to BWI/DCA).
1989
jLI III
-153-
SUMMARY OF FINDINGS
O-D routings with PeoplExpress in 1986 were almost
exclusively through EWR, and PE had an average of 8%
MS in 20 selected markets.
After withdrawal of PeoplExpress from EWR:
-
CO became an effective competitor in many of
these markets, but traffic was split between EWR,
CLE, DEN, and IAH hubs.
-
Growth of alternative and new hubs operated by
other carriers further reduced attractiveness of
EWR connections.
-154-
BOS
CLE
DEN
PITEW
LAX
IAH
O-D Routings
After PeoplExpress 1989
(via Continental Airlines)
-155-
BOS
DTW
ORD
PIT
EWR
IAD
LAX
RDU
DFW
IAH
O-D Routings
After PeoplExpress 1989
(Other Carriers)
MCO
1114
-k-
"i , -
V~
j4
-156-
CONCLUSIONS
e
Withdrawal of PeoplExpress has had significant negative
impact on connecting traffic levels at Newark.
-
Continental took over from PeoplExpress, and Newark
(EWR) became one of the 4 hubs operated by Continental
with CLE, DEN, IAH.
-
CO now serves many O-D pairs through it alternative
hubs, providing a bigger choice of departures and more
direct routings.
-
CO did not replace PE as a competitor, its replaced PE as
the hub operator of EWR.
-
Development of existing and new hubs by other carriers
captured additional EWR market share.
*IIIhI.,,,ihMEIhImUI,
-157-
Airline Seat Inventory Control
For Group Passenger Demand
Presented by
Peter Belobaba
Tom Svrcek
May 1991
-158-
Individual Passenger Seat Inventory Control
Assumes Demand For Each Individual Fare Class Is
Independent And Normally Distributed.
B Class
Y Class
Q Class
M Class
Definition
Expected Marginal Revenue (EMR) Of An Additional Seat
Allocated To A Particular Fare Class Is
EMR(i) =
Fare Class Revenue * Probability of Selling Seat i.
15
15
) = .05
P(X > 0) = .999
P( X > 25
EMR(1) = $500
EMR(26) = $25
-159-
Individual Passenger Seat Inventory Control
Example : Setup
Total Fare Classes:
Aircraft Capacity :
4
100
Demand
Mean
Demand
Stdev
Average
Revenue
Y
14
5
380
B
12
6
320
M
35
10
270
42
12
220
Fare
Class
Q
-160-
Individual Passenger Seat Inventory Control
Results
Seat
No.
Highest
EMR
312.78
309.39
304.77
298.65
290.79
280.99
269.16
255.22
239.24
221.34
201.81
181.12
160.00
138.88
118.19
98.66
269.90
269.87
269.82
269.76
269.68
269.57
269.43
269.25
269.00
268.69
268.29
267.77
267.10
266.26
265.21
263.91
219.92
219.90
219.87
219.83
219.78
219.73
219.65
219.56
219.44
219.30
219.19
218.89
218.61
218.26
217.83
217.32
378.94
378.16
376.86
374.74
371.43
366.43
359.24
349.31
336.20
319.63
312.78
309.39
304.77
299.47
298.65
290.79
'Fare Class Allocations
Y
16
B
13
M
34
Q
37
158.48
156.37
152.17
-161-
Group Passenger Demand
Why Is Group Demand Different From Individual
Passenger Demand ?
-
Group Demand Is Realized Many Months In Advance
Examples : Rose Bowl, Mardi Gras ...
-
Groups Negotiate For A Lower Fare
(Bulk Pricing)
-
Unused Bookings Are Absent From Seat Inventory For Months,
Potentially Displacing Individual Passengers
Cancellation Penalties Often Difficult To Enforce
Due To Competitive Environment
-162-
Problem Statement
Given We Receive A Request For A Group Request Of Size S
For A Specific Origin/Destination And Date.
What Is The Minimum Group Fare An Airline Should Charge
Given That We May Potentially Displace S Individual
Passengers ?
"Answer:
Total Expected Revenue Of
Displaced Individual Pax
Min. Group Fare
=
Size Of Group Request
Example:
$ 2,200
-------20
$ 110
Per Group Pax
,.,111011,
1
-163-
Group Passenger Seat Inventory Control
Two Solution Methodologies
Case 1:
Assume Group Is Indivisible. Find The Itinerary With The
Smallest Displacement Cost Of Individual Passengers.
Case 2:
Relax Indivisibility Constraint. Find Optimal Split Over
N Possible Alternatives For Each Group Request.
-164-
Group Passenger Seat Inventory Control
Large Hub and Spoke Networks Operated by Today's Major
Carriers Allow for Several Different Routings (with
Similar Departure and Arrival Times) For Many Origin Destination Pairs.
For Example, Delta Air Lines Service Between:
New York (EWR/LGA/JFK) and Seattle (SEA)
Dept
5:2OaE
7:05aJ
8:15aL
8:20aE
9:3OaL
9:50aE
11:0OaJ
11:29aL
11:55aE
3:29pL
3:29pL
4:15pE
5:l0pE
5:2OpJ
6:45pL
6:5OpE
Arr
11:35a
12 :3 0p
1:4 5 p
1:45p
2:45p
2 :4 5 p
8 :2 3 p
5:10p
5:10p
8:25p
10:40p
8:25p
12:25a
10: 4 0 p
1:33a
1:33a
Flts
377/ 835
1429 / 1655
467/ 233
281 / 233
937 / 623
583 / 623
1601 / 301
983/ 833
887/ 833
1187 / 367
1187
1038 / 367
237/ 300
1425 / 1187
729/ 625
1421 / 625
Stps/Via
ATL
SLC
DFW
DFW
CVG
CVG
MCO
DFW
DFW
CVG
2
CVG
LAX
SLC
ATL
SLC
,
14111
mi,,
-165-
Group Passenger Seat Inventory Control
Numerical Example: Setup
Dept Date : 12 JUL 91
Group Size: 15
Published Fares for
Possible Outbound Itineraries
EWR/SEA on 12 JUL 91
DL
DL
DL
DL
583 EWR
623
99 EWR
197
950A
108P
340P
652P
CVG
SEA
ATL
SEA
1142A
245P
640P
910P
72S
72S
757
757
DL
DL
887 EWR
833
1155A
312P
DFW
SEA
226P
510P
72S
72S
DL
DL
281 EWR
233
820A
1152A
DFW
SEA
1055A
145P
72S
72S
Y
$642.00
O/W
B
$425.00
O/W
M
$325.50
O/W
Q
$277.00
O/W
-166-
Group Passenger Seat Inventory Control
Numerical Example:
Results
Itinerary #1
Seat
Flt 583
Fit 623.
Itin #1
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
24.40
24.24
97.20
121.60
93.80
93.10
118.04
2.59
20.83
19.53
19.41
92.99
19.25
17.72
81.42
78.50
117.31
114.58
109.71
103.29
102.92
100.67
96.22
6.21
77.91
94.12
14299
77.55"
14.92
74.68
74.19
73.15
6781
92.54
89.60
24.21
12058
12.42
11.79
11.51
10.23
Legi1
88.88
83.76'
83.51
66.82
65.29
Leg 2
86.77
85.57
79.60
78.33
75.52
Total
Min. Group Fare Calculation => 1426.75
15
=
$95.12..
-167-
Group Passenger Seat Inventory Control
Group Booking Model Output
Rank
Itin
Out
2
3
4
1)
3)
2)
Min. Group
Displacement Cost
Fare
Estimate Per Passenger
Request for
Outbound
15 Pax
Total
Leg 2
Leg 1
$87. 80:.
.8 87.80
7 ..
.........
4)
$95.12
95.12
78.64
16.48
$137.93
137.93
137.93
0.00
$236.51
236.51
157.13
79.38
Lowest Published Fare
for EWR/SEA on 12JUL91
$277.00
Minimum Group Fare
Negotiation Does The Rest !
$87.80
-168-
Group Passenger Seat Inventory Control
What Is The Optimal Reduced Fare?
For The Carrier:
$277.00
For The Group:
$87.80
Competitive Advantage
Carrier Implementing Displacement Cost Strategy Has
$277.00 - $87.80 = $189.20
Of "Competitive Leverage"
Ilk
-169-
Group Passenger Seat Inventory Control
Case 2: Relaxation Of Indivisible Groun Constraint
Group
Seat
Itin 1
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
Itin 2
Itin 3
Itin 4
259.11
255.68
247.17
246.32
245.43
244.42
234.76
232.73
230.30
228.77
226.53
226.22
224.97
224.46
220.83
139.99
139.97
139.95
139.92
139.85
139.73
139.52
139.16
138.55
138.44
137.59
136.08
135.31
133.84
131.09
104.97
102.82
99.52
96.09
92.52
88.85
86.32
85.08
84.85
84.46
82.47
81.23
77.34
77.05
73.42
Model Output
Optimal Split Over All Itineraries
Itinerary 1
Itinerary 4
5 Pax
10 Pax
Minimum Group Fare (Divided)
Minimum Group Fare (Undivided)
$81.79
S87.80
-170-
Group Passenger Seat Inventory Control
Question:
Why Are Groups Different From Traditional
Bulk Pricing ?
Answer:
In Bulk Pricing, Marginal Cost Of Each
Additional Item Is Non-Increasing.
Example:
6 Bagels at $ 0.40 I item
24 Bagels at $ 0.30 I item
But:
6 group pax at $ 175.00 I pax
24 group pax at $ 189.00 I pax
Each Additional Passenger We Displace Has A Higher
Expected Marginal Revenue Than The Previous One.
The Larger The Group, The Higher The Average Fare
-171-
Group Passenger Seat Inventory Control
User Optimal Strategy
- Be Flexible In Times / Dates
- Be Willing To Split Up
- Book Only As Many Seats As You Need
Carrier Optimal Strategy
- Find "Minimum Displacement" Seats For Each
Requested Itinerary
- Try To "Split" Groups When Possible
- Book Only "Genuine" Seats
-172-
Conclusions
-
Minimum Group Fare Based On Displacement Of
Individual Passengers
-
No Distribution Assumptions Necessary For
Group Passenger Demand
-
Given N Outbound Itineraries And R Return
Itineraries, We Can Find The Best Of N * R
Possible Combinations
-
Optimal Mix Of Divisible Group is No More
"Difficult". All Necessary Information Exists !
-
Better Utilization Of Excess Capacity Means
Greater Revenue Potential For Airlines
Ww''I
,,111
01,
-173FLIGHT TRANSPORTATION
schid)1mag
LABORATORY,
tt$
Wt 1@-
Presentation at the
MIT/FTL -Industry Cooperative Research Program Review
May 23 /24, 1991
Professor Robert W. Simpson
MIT
-174FLIGHT TRANSPORTATION
LABORATORY,
MIT
-- N
Problem Statement
GIVEN:
1. A fixed schedule of flights F for one type of aircraft
- a flight is one or more flight legs
- arrival / departure times are fixed
- schedule is cyclic over a day or week, C
- schedule remains in effect over planning horizon,H
2. A set of crew bases B where a number of crews NB
are domiciled to fly this type of aircraft
FIND:
the cheapest set of work schedules, or "bidlines" b
for these crews during H which does not violate
work rules imposed by regulations or airline/union
agreements;
- a crew trip t consists of a series of flights to be
flown starting from base and returning within
one or more days
- a work schedule b is a set of trips away from base
on various days of the planning horizon, H
-175FLIGHT TRANSPORTATION
LABORATORY,
MIT
Typical Crew Work Rules - 1
1. Regulatory Rules
(imposed by civil aviation authorities for safety)
- Maximum Daily Flight Hours
- Maximum Weekly (or 7 day period) Flight Hours
- Maximum Monthly Flight Hours
-
Maximum Duty Hours (duty time is time without rest)
-
Minimum Off-Duty Interval
Note- Crew trips and bidlines which conform to these rules will be called
legal or feasible.
These rules limit crew utilization to be substantially less than that
expected by airlines from their aircraft, and mean that crews and
aircraft cannot remain together during trips away from base. It is
desirable to estimate the minimum number of crews required to
cover one cycle of a given schedule as it would give a lower bound
on the number of crew trips which must be generated. It is easy to
compute the maximum number of airborne crews, but due to these
constraints it is less than the minimum required crews.
Due to the aircraft flying perhaps 18 hours per day, and a daily duty
limit of 12 or 14 hours, some crews must start their duty in the
middle of the day to cover late night flights. Due to the minimum
off-duty interval of 8-10 hours, crews on late night flights cannot
start flying on the earliest flights the next morning
1--
-176FLIGHT TRANSPORTATION
LABORATORY,
Or
Basis for Crew Costs
There are two kinds of crews: cockpit and cabin.
The cockpit crew flies together for one month, paired
differently each month. Each aircraft requires a fixed crew.
The cabin crew complement has a minimum, but higher loads
causes more members on certain legs. Changing reservation
information can change work schedules dynamically.
There are three components which determine the monthly pay
of crew members at a US Airline:
1. Monthly Base Pay - independent of hours flown
- depends on grade and longevity
2. Hourly Flight Pay - $ per flight hour
- depends on aircraft type
3. Trip Credit Pay - $ per trip away from base
- depends on details of trip itinerary
- may be zero
4. Overnight Costs - costs of meals,food, and transport
to overnight crew away from base
MIT
". IAll
ii,
-177FLIGHT TRANSPORTATION
LABORATORY,
MIT
'N
Typical Crew Work Rules -
2
2. Airline/Union Trip Agreements
- Daily Flight Guarantee (eg. min. hours if called to duty)
- Flight/Duty Ratio Guarantee (eg. flight/duty time > 0.5)
- Flight/ Trip Ratio Guarantee (eg. flight/trip time > 0.25)
- Maximum No. of Daily Landings
- Deadhead Time is Flight Time
Note- These rules may cause a "penalty" to the airline in the form of
extra pay and hourly credit to be assigned to a particular crew trip
if it violates them. The total flight hours paid in a crew schedule
may exceed the number of hours flown in the aircraft schedule.
Deadheading is flying the crew as passengers to/from base
to other stations where their flying begins or ends.
1%
-178FLIGHT TRANSPORTATION
LABORATORY,
Typical Crew Work Rules - 3
3. Airline/Union Bidline Agreements
- Max. Monthly Flight Hours
- Min. Monthly Flight Hours
- Min. Days Off per month
- Min. Weekends Off per Month
- Max. Duty Hours per Week
- Min. Off Duty Time at Base
- Max. Percentage for Reserve Crew Bidlines
These rules affect the monthly pattern of work for crews but
generally do not cause extra costs. Whereas an aircraft may
fly 300. hours per month, crews are limited to less than 100, so
there are 3-5 times as many crews as aircraft.
Note- Due to schedule deviations caused by weather, crew sickness,
or aircraft equipment failures, reserve crews are given bidlines
which mainly consist of periods when they are "On-Call" and
must be able to report for duty within 1 or 2 hours. There may
be a few flights actually scheduled into a reserve bidline, caused
perhaps by holidays or schedule changes.
MIT
-179FLIGHT TRANSPORTATION
LABORATORY,
MIT
The Current Airline Crew Scheduling Process
Stage 1. - Generation of Feasible Crew Trips from Bases
Stage 2. - Selection of "Optimal"Trips from Bases
Stage 3. - Construction of Crew Bidlines for Bases
Stage 4. - Construction of Reserve Crew Bidlines for Bases
Stage 5. - Execution of Crew Bidding Process
Note: 1. It is a sequential, heuristic Process and is not optimal, even if some
of the stages are done optimally.
2. There should be some feedback of crew scheduling problems into
the aircraft scheduling and airline market service planning. At
present, this feedback does not exist since crew scheduling is done
by airline flight operations personnel late in the airline schedule
planning process. There is a need for some early assessment of
crew scheduling problems in airline schedule development.
3. The availability and continual use of reserve crews affects the
desirability of detailled optimal planning of fixed monthly bidlines.
4. A related process is crew re-scheduling by flight operations
personnel when deviations from schedule plan are occuring.
There is a need for good methods of solving real time, operational
crew scheduling problems to minimize additional costs from
disruptions.
MEMEMOV
-180FLIGHT TRANSPORTATION
LABORATORY,
MIT
WA.
Stage 1 - Generate Feasible Crew Trips
STEP 1 - Establish "Flights", F
For various reasons, it may be desirable to have an unbroken
sequence of flights or flight legs; ie., there may be some arbitrary
specification of where crew connections can be made. Even though
their flight number may change for marketing reasons, here these
sequences will be called flights, f, belonging to a set F. Every crew
trip t will now consist of a sequence of these flights.
STEP 2 - Generate feasible (or legal) Crew "Trips", T
Since there are a number of necessary and desirable attributes for
a crew trip, it is necessary to generate each trip individually. It is
not possible to create a crew "circulation flow". The number of
feasible crew trips may be of the order of a million for a typical US
domestic fleet of 100 aircraft, and in the next step,( the selection of
the best trips to "cover" the schedule), the solution may only involve
a set of trips of the order of twice the number of aircraft in the fleet,
ie., we are looking for the best 200 crew trips. Furthermore, there
may not be much difference between the top 1000 solutions. It is
desirable to find some ''efficient'' way to generate only the "best trips"
as top candidates for a "cover" or solution.
Thus, it is vital to find some new way to generate trips which:
1)
2)
3)
4)
have zero trip penalty costs and good crew utilization
start from a given crew base after a specified start time
involve a specified flight or combination of flights
overnight at a specified location
WWI
Ib
4111111111
Ili
-181FLIGHT TRANSPORTATION
LABORATORY,
MIT
Stage 2 - Select a Set of Crew Trips for each Base
1. All flights in the aircraft schedule must be covered by the selected
set of crew trips.
2. Since each trip starts at a given crew base, the flying assigned to
that crew base by selecting a set of trips must be proportional to
the number of crews domiciled at that base.
3. The Selection Problem takes two mathematical forms:
a) The Set Covering Version;
Minimize
[C. T] given constraints
~E.T21
H.T > B
where E is a zero-one matrix where columns j correspond to
possible trips, and have a one in rows if the trip uses
the flight corresponding to that row
where H is a matrix of flight hours per trip, and B is the total
number of flight hours desired to be assigned to a crew
base corresponding to that row
where T is a zero-one row variable to select trip j such as
to minimize costs
Since the constraints allow the row sum to be greater than unity,
deadheading is allowed and the costs include all components.
b) The Set Partitioning Version;
In this form, no deadheading is allowed and the constraints
are equalities. The costs may be reduced to only the penalty
costs associated with the guarantees.
-182FLIGHT TRANSPORTATION
LABORATORY,
MIT
Example of Trip Selection
Trips, t
10
1 1 0
1 1 0 0 0
CIIlll
C c21
Flights, fi
fi
f2
f3
f4
........ = Cost =C2 +C3+ c5+ c6
010
=1
.
0 1
1 0 1
0 0
0 0
01 1
0 1
=1
=1
1
01 1 10
0 1 0
f5
1
f6 1 0
f7 01 1 01 1 1
f8 0 1 0
a
6
8
5
Bases, B
7
4
b
6
16
C
11 0
0 0
.
.
=1
=1
=1
1
1 0
718
.........
........
6 ...........
..........
=1
=1
>
10
< 15
<5
Cheapest solution to this Set Partitioning Problem is the set of trips (2,3,5,6)
With a large number of rows and columns, this problem is very difficult to
solve exactly. With a few hundred rows and columns, there are a number
of interesting ways to get solutions. If the lowest cost columns can be
produced easily, and the lowest cost column which provides needed cover
could be generated, good solutions may be found quite quickly.
The trip characteristics which are desirable depend on the bidline constraints
and the number of crew available. It might seem important to generate trips
which do as much flying as legally possible in a duty period, but this would just
mean more days off per month for each crew. It is always important to avoid
incurring penalties from the guarantees.
"
WWIIHN111111910N
IN,
-183FLIGHT TRANSPORTATION
LABORATORY,
MIT
The Crew Tree -
A New Method for Constructing Crew Trips
It is possible to create methods which generate any crew trip from a
given base and evaluate it for feasibility and cost. Such methods may
be controllable by the analyst in creating new trips with particular
characteristics which can be added to the cover matrix as desired to
obtain better solutions.
An efficient method of finding "best" crew trips from a base is to create
a labelling method which constructs a "crew tree" on the Schedule Map
for the aircraft. This tree is rooted in the departures from that base,
and finds the best crew routing for any flight in the schedule if it were to
be flown by a crew from that base. The definition of "best" can be varied
but maximizing the flight time achieved is a good basis.
The tree stops whenever the daily limits of flight and duty time are
reached, so that it describes the "scope" of feasible crew routings from
that base in one duty period. The labels indicate the routing used to
reach any flight and the starting departure from base.
Whenever a crew routing returns to its base on some arrival flight
a "best" crew trip has been found. The crew can go off-duty at that time.
The analyst knows that for that pair of departure-arrival flights at this
crew base a crew trip has been found which maximizes the amount of
flying achieved. It is possible to extend the tree construction to find the
second best and third best trips at the same time.
Crew Trees can be constructed for each base. Best trips can be extracted
and the next tree constructed to generate more trips. It seems possible
to generate a Crew circulation for one base at a time if needed.
-184FLIGHT TRANSPORTATION
LABORATORY,
MIT
Handling Overnight Trips
For discussion purposes, assume that there is a daily cycle in the
aircraft schedule. Since there is usually a small number of crew bases
and the aircraft schedule requires flights into. secondary cities later in
the evening with an early departure the next morning, there are
identifiable "overnight" visits for aircraft and crew. Since these
overnight visits cause out of pocket cost, they require special handling.
There may be more than one crew overnighting at certain cities.
The crew tree will show the "best" way to route a crew from any
base into the overnight arrival flights (if it is possible). Since there
will be crew duties starting the next day at these bases, A crew tree is
constructed from these overnight bases showing the best way to route
crews back to base. By examination, it is easy to find the best two day
trip for overnight crews. The search can be extended to three day trips
if it is allowable or desirable.
The selection of low cost, efficient overnight trips can be made first.
Once they are fixed, then all other trips must start and end at their
crew bases within one day. The departure and arrival flights used for
overnighting are then removed from the Schedule Map before
constructing the one day trips.
The crew tree method is a new way to generate candidate trips for the
second step of selection using some search methods of solving the set
covering or partitioning problem. It is designed to only put forward
the best candidates and keep the selection matrix very small. The
process is not optimal, and it is intended that the analyst should be
able to participate interactively in these searches, and return to this
stage after the monthly bidlines have been initially constructed.
11
hil,
, , ., , , 11,11dibl,
WillivilliNIiiiijiillwk
U1
114
INA11,
-185FLIGHT TRANSPORTATION
LABORATORY,
MIT
Stage 2 - Constructing the Bidlines
Given the trips that are to be used for constructing the bidlines, there
are a variety of techniques to generate potential bidlines which obey
all or most of the bidline rules. These rules may be soft in the sense
that crew schedulers may know where and how often they can be bent.
There are research problems in beginning and ending the monthly
bidlines, or"transitioning" between months. For domestic schedules,
there are also problems arising from weekend deviations in the daily
schedule. These problems may be handled by Reserve Crews, but if
good methods of constructing bidlines can be automated, there are
likely to be efficiencies in the number of reserve bidlines used (and
therefore the crews required to support the fleet).
One method used to generate bidlines is to create an efficient "pattern"
of trips over 7 days, and to involve 7 crews in flying exactly the same
bidline for the month. This reduces the size of the selection matrix. A
much smaller matrix of trips versus patterns is used to select the "best"
set of patterns to be used. The focus then changes to finding good
candidate patterns for the bidlines. The patterns can be "mixed" to
provide some variety in the crews' monthly work if desired.
The solution of the bidline selection process once again requires a
good heuristic search methods of quickly"solving" set covering and
set partitioning problems.
q%=
-186FLIGHT TRANSPORTATION
LABORATORY,
MIT
Conclusions
1. There are some new approaches to creating interactive methods
to support the crew scheduler in finding good low cost schedules
for the crews.
2. There is a need to create methods for re-scheduling crews when
deviations occur in executing the schedule. This should affect the
current use of reserve crews
3. There seems to be a need to create similar methods for the cabin
crews which are responsive to their differences in scheduling rules.
4. There is a need to provide some "early warning" methods for
market and aircraft schedulers to cause a feedback of expensive
crew scheduling problems before the aircraft schedule is finalized.
SCHEDULING SYSTEMS:
COMPUTER AIDS FOR EXECUTION RESCHEDULING
1.
2.
3.
4.
5.
6.
7.
Execution Rescheduling
The Influence of Rapid Advances in Computer Technology
The "Airline Scheduling Workstation" (ASW)
A 2-Stage Development Approach for An ASW
STAGE 1: A Manual, Interactive Graphics Scheduling System
STAGE 2: Automated Decision Support
Conclusions
Dennis F.X. Mathaisel
Flight Transportation Laboratory
Massachusetts Institute of Technology
Cambridge, Massachusetts 02139 USA
(617) 253-1761
-188-
1.EXECUTION RESCHEDULING
GOAL:
- execute the operational schedule at least extra "cost" due to
schedule aberrations
INPUTS:
-
Operational schedule
Operational deviations
- Weather, breakdowns
- Late arrivals
-
Expected traffic loads and revenues
Short term operating costs
OUTPUTS:
- Modified execution schedule
- Cancellations
- Delays
-189-
II.The Influence of Rapid Advances in Computer Technology
-
"Techniques in search of an Application"
- use mainframe: large, fast supercomputer
- construct fixed code for technique
- user submits data, receives solution
- user reviews solution to comprehend it
- causality: user cannot ask for explanation of
solution
- user may interface with OR analyst
New Approach
-
"Customize techniques to the Application"
smaller, interactive graphic workstations on
-
create various fast heuristics to solve
-
create links to solve large scale problems on
-
user ismaster, computer isservant, direct
-
processes are custom built to meet
-
systems to match existing procedures and
common network
subproblems
mainframe
interface
application needs
organization
-190-
Ill. THE AIRLINE SCHEDULING WORKSTATION (ASW)
A COMPUTER TOOL FOR AIRLINE SCHEDULERS BASED ON THREE NEW
TECHNOLOGIES:
1. Table top Engineering Workstations with a speed of 1-4
mips and disk storage of 100 -1000 MB working together on a
local area network, interfaced with existing airline mainframe
systems.
2. Large (19 inch), high-quality color displays with
interactive,instantaneous, manipulation of schedule
graphics information using a "mouse".
3. Object-oriented programming to provide modular code,
easily extendable to handle time-varying
scheduling constraints, policies, etc., and to reduce
programming support.
We shall call this tool the ASW (Airline Scheduling Workstation)
-191-
IV.DEVELOPMENT APPROACH FOR AN ASW
GENERAL DEVELOPMENT STRATEGIES
- Involve schedulers at all development stages-- (there will
be cultural and organizational shock)
- provide familiar systems and reports first to ensure that the
new system will not preclude doing certain schedule subprocesses by old methods.
- Expect changes in organization. and procedures as
workstation capabilities are perceived.
- Establish a local area network of workstations in scheduling
area, capable of interfacing with the airline's existing
mainframe system. (e.g., 3 workstations at $15,000 each).
(Establish a "Schedule Generation" workroom).
- Develop modern, transportable, modular, object-oriented
software, for automation of sub-processes in scheduling
- easily extendable
- easily supported
- C, PASCAL, LISP language
- good data structures
- A Two-Stag e development process
- TAGE 1: introduction of manual, interactive
graphics scheduling system
introduction of automated decision
support
- STAGE 2:
-5-
-192-
V. STAGE 1 - A MANUAL, INTERACTIVE GRAPHICS SCHEDULING SYSTEM
A) Provide computer graphic displays of schedule information
- instantaneously modifiable by mouse
- global data base modification
- selectable screen data -- by fleet, station
- save alternate solutions
- audit trail
- memo pad for scheduler
- keyed to input data, and assumptions used
- automated search routines, etc. to minimize keyboard
and mouse work
B) Provide instantaneous error flagging (even if error occurs
off-screen)
- e.g., insufficient gates, flow imbalance, double crew
layover, violation of turnaround or transit times, insufficient
aircraft.
C) Integrate crew, gate, maintenance schedule with aircraft
schedule
D) Provide familiar printed reports and graphics for distribution
around airline
E) Provide interface to mainframe data system to maintain
current scheduling processes.
___I
OTHERS
.. -.......
REPORTS
EDIT
FILES
I
INDOW
SEARC
INTERMEDIATE
FUTURE
WEEKDAY
]
QUIT
ALGORITHM
___I
4W
OTHERS
REPORTS
EDIT
SEARCH
FILE
.-
~...
WINDOW
R.OT A1T1ON_ .CHA RIT
SCHEDULELISTING
STATION CHART
STATION PROFILE
SOB PK0F4IE$
I
EKDAY
QUIT
ALGORITHM
00
OF
:
M2
ne
OF LAX
NR
is
SEA:
:2st
LAX
fe
676 B(
PSI
NCO
: 221
ts
ORD
PBI
CO
-
MIA:
26s
Al
FLL
to;
96349/84
2820
SFO1
a
:
ATL
146
LGs
L AX
BOS
s:
;.
a
;1
494
ATL
|PI
|
8820
eoeo
t39
ATL
NCO: 12
t7 :
ATL
a
s
X
8998
7
LG A:
ufLouf~p
*
FLL
6s
t26
ENR
364
FL
U434 |
2"
SDL
miamagp
44s
|ATL
;ese9
CO
FLL it
ON001=
NCO:
too
DFN
11
a
1 I
"
JFK
LGA
1912
Los
a
FLL
L10
0
ATL
|
a2455
a
63
BOSa
'us
8~56
egas
1180
*
e63497e4
MCO:
22
7
S.
66
ATL
3s4:
G~54
04,54
LIO
Lie
OF
i
208
010301100
1s
,
TPA
221
S6
#:
0091
14
13
SF0
2
BOS
108
I
2
1
to
e9
ENR
so
ATL
us
006
LIO
es
I
SEA
2010
LID
er
es
LAX
:261
ATL
es
An.
s'
DFH:
002
L10
2 0 4
e0 0
5.
Monday
deltanew
ROTATION CHART
GMT~e
istF[
3ss Al
MIA
ess
DFN
14s
F
FLL
8B4
16
ATL :us
: M(
9829 8898
es4
105
n2
JFK:
66
ATL El
FLL
2"
016
NCO
t2:
ATL
fss
TPA
EMR
0850
zu
FL
ist
9'
2 105
mi
I
EDIT
ALGORITHM
ADD
COPY
delta ni
ROTATION CHART
DFN:
se
002
LI
za
z
s
:LAX
OFN
SEA
ii
ENR
as
TPA
DI
SF073s3
MO VE 1
MOVE ALL
so
SF0 O
A"l
13431
SEA:
SWAP
LAX
Ilre
UNDEi.ETE
ATI
:
:2s
A
3
MCO 2f
006
MIA:
L10
I
-
--
14Z6
:ATL
:
6s:
:
n
FLL
013
LI9
iFLL
ENR
364
2125
I
DFN
FLL
MCO:
tse
~yu~uysi
tstF1
SF
|
das Al
MIA
44
: 49
is MCO
b
U51 !q4
e44p
:PO
JFK:
ATL
tz
s
AlTi
L
FLL
:
MCO
* ui~u~'~p
TbY
2 BL
se
17 |
JKi
22
1in
ATL
U~34 I~bI34
S
*
to
ATL
134
|PBI
LGA
is
:12
1q12
ATL
s3,
FLL
2
0
BOS:
LYOb
:66
8p88
LGA:
011~in
434
*
LAX
BOS
~2
10
LGAs
-
S.
88:
'
S
2
LID gl
ss
'Its
JI
SFO:
Al
is
U~34
NCO:
FLL
Is
:
BOS
ORD
ATL
3s4:
Io'
007
:sis B(
PBI
BI
: ATI :
221 its
22
6
005
LID
015~O
LIO 010
15
77-7754
1-04
012
LID92
14
13
12
MODIFY
LAX
003 ggg
008
0
11
10
09
INTERCHANGE
AiTL
se*
Monday
DELETEALL
0
04 2
~8188~
DELETE
21
143
e
ts
si
:
,
:
ATL|
II
UF
is
M(
ATL E)
FLL:
J.
011
feiL
761
MCO
123
ATL
i ss
EHR
TPA
I
08
'ot
'
Fl
ENR
Ii
SLI
I
MIA
NCe6
6139
9214
CC
A
394
of6s34l
45,14
''no
9549
225
:
LAX
Ld
9p
U
.
66
'969
9~4
6899
III
SI
P
ATL
4
LGA
SFO
-FLL
S
36
S
I
ATL
IIII
SI
ISII
II
I
I
I
uso
FLL
MCO
II
I
S
FLL 0
01
0
07 ul
767
se
DFN:
1
DFN
002
03
04
05
s1:
AiL
Is
02
07
10
09
L0
ENR
gg
13
T: iPA
:
: 01
I
1'
1Op56
AlL261NCO
0900
:
a2aMi
tiit
10B
tt.
s
BOS:
_
_ _
_
a
a
a
i-u
El
9
SJU
:3
ATIL 7s
96
Al
ORD
1
:
BDA:
ATL
is
a
a3t apa
LOSAX
(
4
aT6
arw
Al
%6
a
a
a
ATL___:___:______MA_____
pau'.4
tsi:MCO :112
AlL
F
:i:
LAX
Al
t46:
AlL:is%
LAX:
FLL
77
DI
1
__________4_
sas
DF
16f1713
ATL:
a2
a
ao
LGA:
1 51024a
SJU
1155
_ _
Ait
S F O
m
s
008
:
so
ol
_ _
_
NCO
205
01
063fo4
5
:NCO:
a
a
072gsamluaa
:ss
3s ATL
1HIA:
L
ATL
11
_ _
ORD13
s
:u 1
1
_
PBI
u6
I LAX
1
4
1122s
BOS tit
s
2$
AIL
NCiss
14s
1E23
PBI
'
ATLisv
22
21
DFN as
:AX
a
20
1'~
so
:
a
9
1o8
as'
ie
I SEA:01
132
1170
17
16
!~
SF0
2oIo
007
I5
14
2
005
Oil___00__4______
FLL
012
118
0s
11LAX
JFK
LGA
a6,
is:
DFN
MCO
~:
:
ts
aL
1
a
a aBD
015go
JFK:
n21
a
aa
asMOu
a
Fl
151
:AlL its
a
PI
a13a2d~~
AT
I
I
:NCO
ag
1a
JFK
144
Al: 7s
MIA s
Al :443
a
FLI :176
L1
110
a
:ELLamCO
01L
04
364
ELI
ENR
12s
is01
0
12
z17
SEA
16t:
ATL
010
11
ENR
LI5
004
ALGORITHM
QUIT
WEEKDAY
Monday
08
:LAX
:LAX:1-43 n
003__
0Q
I
:s2:
WINDOW
FILES
SEARCH
delta-new
ROTATION CHART
GMT
EDIT
REPORTS
OTHERS
isi
4 Al:
1.2
A
si
so aAlL:
:1221! 52
IBDA:
;IIl
Al
FLL
NCO
62'
:I0Ss
p
:14
76t :NCO
123'
ATL
ass
TPA
a
0400
a1
2a
a
Ep0
R
*b
FLL
69
ENR:829:
i
0
: ELI
ENR
2
16
~
A
i
t
05
FL
:Al
saJF
OTHERS
REPOR"
delt a n
ROTATION CHART
so
L02 011
: DFH:
:LAX
Monday
Monday
DELETE
DELETEALL
INTERCHANGE
TPA
asa
ATl
tie
SEA:
L04
2 1A
1
2
2
NCO
aI
g
576
: OS
:ORD is
AlL:
I
23
2,127
ATL:
15o
a
FH
I
I
91
tit: NCO tit ATL
ll :s
22
M:IA:
F
~
ATLU6PI
NCO:
L
2C2
0080
iLs A
LAX
ri
96!
B:OA:
61
Al
66
:S
a :
a
:
:i5
:MIA: 97
ATll
:
LAX
:Lu
Al isi:NCO :tz AlL :
%6
as
aF
LAX:
Lm
NOS
IGA:n
AL:34hl:
ae
TLt6TA
:FLI
0a
:134
.
a
a
4
:
-MIA
m21f
aTANa
085
aw
Al
UL
is%
Al:
F1
isi
AT:t2NCO
14
: In
.FLuM4
____87______6s_:__
a
B
t 46:
a,
a
ORD
63P
;
AT 739: P11
ins
:Al
?4
JFK
144:
aLLlaC
a
78:
l
aT
C2AL2
a
a
a
ape
a
a
F
7t
110
:14:
ATL
Al
:32
C0
011 00100
015 1
:10:
Use
SJU
:
66s
:494ATL PBI
LG6
2
: EE N:
22 7
10
I3 010
'
FLI
2
AL
394
:Lf
t~
SJU
:33:
OF
tit
L05
106
24
o DF
LAX
:s
a a
I
LLUb
ZOM:PB1
st NCO
2st ATL
119
is
:LAX :t~s
LINDE LETE
EHR
107 1
DFN:
SF
MOVE ALL
MODIFY
SWAP
SEA
70
ALGORITHM
QUIT
WEEKDAY
WINDOW
FILES
RCH
ADD
COPY
07
:z : LAX:
DFN
L03 000'
ENR
:
Ail 03
,
I1
1
EDIT
I'~
a7
1~a
REPORTS
OTHERS
oo
-7
0
1 so DF: sos
DFH
AL
07
50
08
Monday
0 :9
10
L0
03
ENR
0
L
t
SEA:
ORD
MIA A3s4 ATL
NCO2si
L0 722
10a
FLL
2
LAX
OS
ses
n
s
: ~a
:
ATL
:194
2756 0eyUWv4
pyIu
JFK
:JFK:
9
Al
ORD
7:
B(
?4
ATL
os
ATL :
s
l
LAX
ul
134'
:FLL
:5t :
FLL :364 ATL
too
MCO
:44s
sas
s
L(
j
: Ft
i
PBI
14
A:TL t@2
176
: I
ATL :u1s : NCO so ATL
ATL L s4
ENR
:
S
a
a:E
obt
ENR
FLL
BDA:
es
aO85a11~l
~u
ATL :
ENR :25'
.".5'1~qu~'q
I
o :JF
FLL
:87
Us :
U0Ssa
62'
:
a
a
MCO
SFO
4:
:gi 01 2 a'2
: :1
~
Al
-JFK
a
ATL fss TPA
2-
t46
ATL:
:
144
:ATL iss:
DFN
ts
F49LL IC
a1geg
FLL
n;i:F F
MIA
2vsoosi
CO 12W
B
if
3
1pe
EDL
761:
MIA
6
a
All TLs:MCO :112 ATL
LGA
e32
e5
BDA:
61
'2
DFN
Lf
s,
ATL
:2'
LAX
126 : ENR
13o:
BOS:
is
FLL
76
05
SJU
66
4s4 ATL PBI
2355
LGA:
0116"U
LGAs
SFO:
L
ATL
itu
sm1
2zgzmmm
90
NCO:
FLL
ot
'
7 BOS
DF
I
aa
2
00006
71
SJU
:03:
ATL:
is
1
5a.24|
2
-
e
s
NCOtit ATL
BOS ut
P:I :67s
ATL :ts: PII:
05G
DFN
to:
LAX:
2
, ATL:
ts
1320:
,134287
II43I
ATL ist MCO
st
NCO
ATlliss
so
SFO
so DF
: LAX
is
:DFN:
24
23
22
21
20
19
1
17
1
15
TWR iPA as
s@ ;ENR:
SEA
n11.1
14
13
12
11
:LAX
:LAX
ALGORITHM
QUIT
WEEKDAY
WINDOW
FILES
SEARCH
delta-new
ROTATION CHART
G-----a
T
EDIT
i-
ALGORITHM
QUIT
WEEKDAY
WINDOW
FILES
SEARCH
EDIT
REPORTS
OTHERS
M
-ROTATION CHART
7
I
delta-new
I
15*
g
2
I
I
DFN:M 9'
002 g og DF
L1
:
A&
SF
SEA:
-
92 : LAX
SEA:
:LAX : nt
:
A
LAX
003
FOTPA :i
EN R
04
EHR
k6 :
=I.1
2T
L u0
2
:
: : SF0:
tit
AT
s
LAX
OF
ATL:
V
D
NCO 1 tit ATl
f.
: OF
AI
s
OS:
a
8 26'
:ATL
:94
BOA:
61
LB MI AWs
a s: A"T
:4s4ATL PBI
:
:
NCO
ORD
:
t 34'
ATL:
BDA:62
86
so'Al:
ICO
FL
NCO
SF0
t 49
143
Is% UL
ATL: 162
14
:Al
:
JFK
144
PBI
ATL 7ss
MIA ins
14
t 46:
LAX
FIL
LL
:FLLaMCO
ATL:.
ATL
si
ILI
:
Lis
6:ENR
:Al
ATL
:s
MIA
60
ATI tsi:NC0 :In ATLl
%6
: 3
:SJU
:
:s3
:66:
:364ATLNR44
t
Os
LAX
FLL
0
1TDFL
:
:LGA;
LAX
71
o
ATI
is
:676
7
IA:W ATL
639F4
10104
2
01801081:
13
I
s
SJU
OO~1gsi :
57
RD
SFTL
ALlseg BOS
GA:
ATL
s@@s9I'
ORD:
NCO261
FLL 7L ANC
tot
lOS
11
08ggiggM~i
I
ATL is1? CO
002sMIA 116 P3A:
L06
I
DFN:
I1524:
I
715
t
ay
PB
I
u
s
iLAX~
tis
A
_
:
-_
*TPA
1
0
'
on
IOSs 6k
D2s
t21:FLL
JFK:
01 61: COZ
01066
LI
2
ATL
"Wsp
ATL fs6TPA
:
2
::
:
:
I
:
I
:ENR:
I
0
obt
:194
FLL
:
NR
es
memas
:
u'
:
FIL
E R:s2s'
q
esy
66 JF
|ATL
,ggl
I
tua a
OTHERS IRE POR
m
w
T
7
n c
es9
As
DELETE
DELETEALL
delta_n
ROTATION CHART
I
A
1
9
2
TPA
NR:
Monday
17
un
L000108'
SEA:
0S
ZOOM
ATL:
10
0010
007ggooio
10
79,
ATL
MIA: s
MCO 261
06
SF :
008
100
ATL
t6
BOS:
ATL
L1G300100
f13ooioo
FLL
ATL
FLL
aATL us
|
NCO s' ATL
ATL
|A1 |
ENR
FLL
6ms
us
6 Oo10
ENR
sat
g80
as__l
~n
JFK
14
PBI
:ATL: 79s
?ib
SEA
ENR
|s2'
gg
it
ATL
2
A TL
ts
PlI
ORD
FLL
tos
BOS
14
S
ses
ATL
BOS
2
LGA
ts
FLL
tss
NCO
FLL
n7
:12s
7s
ENR
W-
LGA
a
BOS sk BDL
|s
JFK: 12a
:NCO 128'
ATL Cs6
2a"
A' FLL
g
st
9
7:
ATL: t@2
4
a
iBDA6',
s
ENR
146
ATL
14
sis
a
LAX
s
1re
ist
:
t
ATL
ATL
9
:LAX
tra
ss
is
MIA
s44
ts
ATL
:
DFN
s s
SJU
LAX
LAX
FLL :364 ATL
aa I
s'
DFN:
o
tit ATL
S5 2
t
JFK
s6 TPA
:
~~
aCO
esag0e
011
O16@gg
: MIA:
6
r ATL
t97MC?
iua
BDA:
63
spee1
L009001001
94
93
1 55
s2
SJU
4S4ATL PII :194
LGA
LAX
;s
its
aCO:
02
01
s:
ATL:
DFN
U LNDEiETE
ORD:
24
23
LAX
MCO its
:
MODIFY
SWAP
26t NCO
L00188
:
s:
ATL iss
e
LIG
003
22
21
MOVE15241
SF0
002go,
20
19
is
DFN
INTERCHANGE
0
ALGORITHM
QUIT
WEEKDAY
WINDOW
FILES
RCH
ADD
COPY
r1111w
:ATL
|76t
gg
FLL
delta-new
ROTATION CHART
MCENR:
01T
002
s
SF0:
ap
0 43I
SEA:
t
:
it:
~
BOS
67s
L10
NCO 2s1
MIA: s
'0
L06
Ibq6
ATL 2st
EHR26L
ats
NCO:
SF0:
007~i
LGAs
4s4 ATLPI
B0S
:s
ATL:s:PI
2
ATL
I:4
is
sSf
2
12
:1
ATL ist:NCO 112 ATL
LAX
LAX
ATI: st
146:
2
015
14
Lo16
JFKQ
Li80
0
FLL a
)
ATL
s
ATL
3
FLL
176 ATL its
N
FENR6
ENR
1:4
ATL
LiO
esATL:
T ATLiss
FLL'
*
sit
085
100154
ts2
-FI
LGA
76
FLL
EMR
:2s
NCO
n FLL
S
I
|NCO
ATL
1
PBI
ss
DF
FLL
LOY
016 80180i
MIA
ti
JFK
:AL-s
14
ILA
FLL
:: tst
BOS
m
__1
93
&pUa
O'e
: FL
i
BOS
24
ATL
ATL
MIA2 s'
6s
ORD
ATL
2,
BDA:
s1
2
:
-4
IU
;
01*
J
SJU
66
ATL
SEA:
i
:LAX,
705
0051601881
10
07O
LAX
,
1715U1
F'1:
ATLf41
SJU
s3
ATL:
is
tt
NCO
ATL
1;u
u1Y
:
05J1
v5
,2
DFN
tr
DFN
: :152
:1~
m
s
DFN:
1 1
s
ATL:
14;55i~;p
31
128
PBI
NCO
ORD:
mCo its
LA
VG HI 3 ~
1-40
03iO
261
ATL iss'
'
:LAX
s
LAX
as:
Monday
04~
a3
82
a
24-
23
22
21
20l
19
in
: DFH: '5
i
TPA
a
*
7
t
a.
EEDAALGORITHM
WIND
FILES
SEARCH
EDIT
REPORTS
OTHERS
I*
BOA
:Fe:'
E4
BUS 6k BOL
62
FLL
s@
FFK:
:463
ff1%
12t,
FLL
AATL 1A5s
N
j 2'EIT
2MM
OTHERS
ROTATION CHART
GMT
05
06
07
08
s
EDIT
REPORTS
SEARCH
10
11
12
14
*
lie
:66A
00
NCO:
24
DF~i
02
01
sin:
l
t~.
tit
:
:IOS:
s1
poo
f
04
ATL
so'
:DFN:
:FH
05
LAX:
s2
SEA:
70A~
I
'
aL1--s
Sa a
fi
26t
ERRLX
I
:
DF
1
AJL
s6
E
a
LA
ATL i:s P
aT1
153
is:
W
J
tr
.a
R
24
ATl
I
:105
Lot
FLL
II
a
2945.Zpe
0 08 @pl@6j -
:
LOS9881883
009s Bew
AL i:NCO|u2ATL U2 AT1
LAX
10
p
a
a
a gs4
a,10
er
vV
LAX
;
LAX
JFk
L1
:
02:
FIL :364 AlL :443
:FLit.MCO
:151:
MIA sis
:
:134'
FLt. 144:
AL: 7ss:
PB1
ATL:
sj
-iFL
FIL
:ATL: t#2: NCO
NCO
sAL is :A:
:XATL:its
ENR
126s
F :s
LS 6
a6
ATL
015
L~~
1
LI Dmmn
'
a
'i
c IPA
a
a
: : : :
a
a
'
ENR:
a
sat
194
ER
: FLL
tes
ml10
8t
amm]
FLL
JFK: i2
F6
EN[R 54:s32'
ggmaiamip'
10
:ATL rut
C
FLDLL
10
010
IGA:
:76
l
s'
BOS
2
tsSF0
FLI1
ses
is% ILA
JFK
14
19D
ellees
211:
ATL
46
eYev
0100:1:0:
0 6a
03
171
ATL
:2'
1
BOIA:
'
NCOt
AW
DF
a
:s:
SJU
:lp
; esagr4
its
132
AT SJU
LID
L
ALGORITHM
1 24:~j :12:15
IA: 3ATL
154
23
ts~vN ts:A
:LAX:
AL
10~
22
LAX
O56
005:
MIAaa
1008
21
:s
PIILAXa s676IOS
113y297
NCO 21
L10
20
19
DFN
is
1
LjO4
2
17
Ao TAL
0:SEA:
003
16G
.15
,,SF0:
LIO
%;E
is
QUIT
WEEKDAY
Monday
I3
TPA
E:R
0,
143
2t
MCO
LIAD
WINDOW
delta-new
09
0
004
FILES
MCO
123'
280 R"A
2F
a
as
FLL
ATL
2P
156
EDIT
ADD
REPOF
OTHERS
07
01 0
,o
deltar
Be_
_
: EN
TPA
R
e
eMODIFY
*
zoomET
ZOOM24
261 MCO
LIO
tPE3 01861
ORD: is
C0
MIA: ws
815,1
26t
8 0000
L00
LGs
SF0:
SJU
.
SJU
66
~
;
uls
NCO:
LID
:
1105
ATL
:1s4
DFM
* 52:~
ATL :7s
:2'
715
ATL
16t
2
1t
ORD
I
is
IDA
s7
MIA
s
r2
u
SEA
na
:LAX:
01
: LAX
2
: ENR
i
ATL
NCO tit 16-1713
9V
ATL its PBI
96
ATL
BOS
1
FLL
ti:
s
ATL
pe13
2 55p
LAX
%6
6
6
|
6:
1
6,up
LAX
6
102~gg
JFK
FLL
:FLL
:364
ATL
CO
:44s
LAX
uW
ATL is|:MCO :ii ATL
ope65
15 o
L0400
DFN
el
o11
I~e
6 !
BOS:
494ATL PBI
tit
BOS
,
ATL
ATL:
*
SWAP
06
NCO its
:"wrli
ATL iss
MOVEALL
SEA
00
0
1
et MOvE1
002 010
At
se'
DFN
so
LAX
,ogFH:es
TERCHNGE
SFO
787
1-0
Monday
DELETE
DELETEALL
12
to_11
9_
ALGORITHM
QUIT
WEEKDAY
WINDOW
FILES
C OPY
O TH E RS__ _ _ _ _
ROTAT ION CHART
-
ARCH
FLL
is
MIA
ATL 799
ess
JFK
144
PBI
4
145:
DF1
f43
t12
SFO
NCO
FLL
LGA
7s
EHR
t26l
DFM
tit
BOS
n7
FLL
iss
ATL:
14
2
LGA:
isw
ATL:
134'
ses
ATL
46
tes
NCO
4202
FLL
6 TPA
ENR
08650
sI ATL:
NCO
s76 ATL is
ATL
|4s
sbt
|
1
FLL
:IDA:
s*
ENR
6es
62:
FLL
s'
EMR
s296
'1OS
BL
so
ATL is6t
JFK:
12i
FLL
NCO iws' ATL is6
20
266
42
EDIT
REPORTS
OTHRS
QUND
FILES
SEARCH
Monday
delta-new
ROTATION CHART
GMT 5
6
7
0 :9 0 1 11 11
14
5
is
19
:DFN:
20
ATL iss
3
10
0050884
s
BOS
06aai
15
LAX
10
W:
BOS
82,
SJU
__
e05s3ge4
FL
0192 10188
I
14
103
ATL
364
Tulv
MIA
1
;H1995
-
t6 ATL us
tOsA LGA:
u
a TPA
I
g
ps0
JFK
14
ae
1:,
ATL: 1e2 NCO
q91
SF0
:ey~5~
NCO so ATl
'
s
:BDA:
:
62'
:ENR:
FLL
n
1
is
FLL
iss
4:
pa1
LGA
:1s6 ENR
m
Ifr
DF
isi
NCO
iso
42m ae mnamnm.
:BOSs6 BL
aa
60
1I5
L3
ATL iss: PBI
ia
*2,
-
4f0a%1
Is
DFN
14:
ELL
10
449
BOS
ses
ATL
:
2
ATL
144
ENR
ATL
|s4
:
sIm
:sm
:FLL
6es
1
Ua
FLL
ltI
SF
1"
11AL4:
FLL
ist
JK
:FLLiMC0
,
*
L*
3
If
BOS
ATL
97
i
I
I 2
1
LAX
1:1
LAX
ATL 2t
-1
26t
ORD
ATL
MIAA
u:
ATL
7s
ATLI
10
o 81881
II
EMR
14i
ATL tis PBI
passumiggi2
55
:07
SEA
n0
;LAX
1e1Pu24: 151713 ;
:
:JU ATL 6
10 n:
t :MC
ATL
%6
:LAX
s2
I
is
BDA
I
PBI
As4 ATL
pe
3
ss
007
LGA
0
715
NCO ii ATL
: iii
SJU
66
l
CO
DFN
524:
: II
0
A :TL :
MIA: 3s4 ATL
MC026s1
o4
ATl
s'
:DFN:
ut
DFN
:p
L0
OR: is
o2 03
1 -
AIL T a
NCO its
:
PBI
L04010 "'
24
s:
LAX
:
Arm
NCO
23
12
LAX: LX
i
22
1s
0~~(43;1327
SEA:
21
1
' 4150
so
gg SFO
*
*
*
to
*
.734
L10
t7
::
TPA is
sI ENR
SFOC
THM
FLL
:v
se
:JFK:
S-Uu
aaIs.g2
ENR
15lig
32'
:ATL: :t NCO
AT
123'
1 umsamiINppNa umu p ;gga
IIbIa
FLL
in2
2
2"
9
8
07
GMTeG
Monday
delta-new
ROTATION CHART
's
TPA
I
ER:
s:
tie
0m
L04
73'
as
ATL
MIA:
NCOs
09
2
%6
LAX
6681
L*
10
4s ATL PBI
LGA
:SFO:
: BOS
:
£s
BOS
:134
ATL
Al
tsiNCO:tizATL
05Y0
,q/W
MIA
65
:OS :
74
ATL
FLL
S0 s ATL2 iss
FI
OLL
t7s ATL
s
JFK
144
Tau
:15
s TPL
:2:
2
ATL: t@2
4
PBI
LGA
FLL: 7s
isi
EMR
12
NCO
5u
NCO s
IU3.5
ATL
0
FLL
iss
'
MC:01
I
1
t3
IGJ
L
L0
BOS
22 M.
2N.1
I
I
L1
ses
ATL:
:46:
0
O12p @8
: FL
n
SF
.
MIA
, LGA:
_.
I
I
ATL it
ORD
I
I
134'
:1ti
iL0e
SEA:
261
ENR
is
ATL:
173~~e
JFK
nt
ATL
7
LAX
1
FLL :364 ATL :441
78:
LAX
0
1 1
is
BDA:
:LAX,
Its PBI
: ATL
:
SJU
:
2
ATL
s
4
ATLm
:LAX:
fi
5
a
s
DFM,
DFN
i7t
NCO tit ATL
tut
SJU
ss
as
NCO:
DFH
s
ATL
19
ATL:
tv
si;
ORD
NCO its
:
s:
LAX
;
:8
ATL iss
LAX
ziNCO:P1I
20t~
:DFN:
I
SFO
SEA:
s
5
14
1a
2
le
82
0
23 24
22
29
2,
9
0
t
ENR
1
s'L
:t4
"
FLL
I
ZI
ENR
|
s?
Iu
SATLF e BDA6F OS6 BDL
FLL
#
n1"
6s s
ENR ;AT
JFK:
69
11199'
1t,
FLL
2V
C' t
Ts76A MO
ATL
C3~
91
I
I
R ATI N C
delta-nI
ROTATION CHART
GMT
05
06
LIO
0s
10
iPA
Monday
DELETEALL
1,2
1
ENR:
700,plw
67
002
e7
09
:DFN:
INTERCHANGE
MOVEi
MOVE ALL
21
20
19
18
17
261
ATL
a
DFN
--
00
0
008
2si
10
'
MIA: 3s
ATL
SJU
A
NCO:
tie
BOS:
LOSS
LGAs
SFO:
4s4ATL PRI :14
s
LAX
us
LGA:
is%
134'
L10
ATL
146:
01301100i
LID
i905 a
:FLLfiMC0
:443
1
ATL
74
lOS
FK
is
:
144
FLL
it
JFK
ATL
SF
1OS
se
193ts
24ags0
LAX
364
FLL
n7
:
s
LAX
t:
010
s
a
DIN
:ATL: 7ss
a
PI
14
49
is
FLL
LGA
:
ATL: te2
NCO
SFO
isi
a26
2
919
FLL
ass
J711W
aj~
MIA
2
1u
2
%s ATL is :MCO :i2 ATL :
FLL
225
ORD
epe:1e
L11680101
ATL ist
26t
;
t16 PBI
ATL
si
MIA
6s
"
SEA
mi
1
ATL:
is
ENR
fnt
ATL
131
BDA:
ATL
py
2WS
:LAX
164u71
ATL
:2.
LI0D
10
008
36
pa
:LAX
s2
1
MCO ithATL
it
BOS
SJU
66
:
ui
r
tiszessu145
ZOOM
ORD
se
DFN
ATL U
NCO its
ss
1 5
MODIF Y
zoo
ATL
:1415
SEA:
L410
| DFN
sa
LAX
s
04
03
02
01
24
23
22
o9'
SFO
ALGORITHM
QUIT
WEEKDAY
WINDOW
FILES
RCH
ADD
COPY
DELETE
REPO R
O TH ERS
225;
2
ts
NCO
ENR
DFH
2
21r,
aLL
L76
NCO so ATL
ATL its
s
100
ATL
56 TP
0160886g
L1O
s444
:ENR:
sh
0ga850ig
BDA:
RFKl
:3a
EaR
a
: FLL
%is
62
:OSs k BDL
6s
FLL
EMR
82s'
ATL :7t
p120s 15-gO
121,
MCO
tra
FLL
ATL
ss
I
ggp
'" A r] 21P1"
A
CH
pO
delt:
NC :
e
Y
:
OARD, s
-10
1
s
its PBI
ATL Al
ATL isA lCO ts
:
DFN
:L AX
z
m
v
M
q1
ROTAION HARTdelt-newMonday
e E 2 as A4 es
23a 24
2 1 22
ATL
2S
:SJU
6TL
s:
SFO
SE
02
17 Is76
I QU
WEEKD A
WINDOW
FILES
SEARCH
EDIT
REPORTS
OTHERS
zetMCO
1
ADF
s6
LAX
s
:
LDA
L
s
sAX AS
05
003
a
00100,
L1
L1
at
00
1
L1
Iw
2a 5
a
a4
a:
em
Iu
MIA
11125
A.
:G:
: Al: 1~
OR
7@
ATL
8si
16:LK
: : is LAI s4 :DN eATL:1: NC
S
FLL 44ATL
Das
IA
14
14''~
12"
a
*6(ws
A:
aPB
$ss
pmma a
a
aI
:'
a53 a
a
a
a
9 a a * 55a a
12
L1
aUl
a0
avo
a
a0
13 tf1C
i
:
aA
14
: Ta :443
364A
144@P
"JF"!%
iHCO sLX
a
i6
1.
IU
1ipeT
uaa
'FLL
l
isM
Jk3FFLoL'
:17s :ATL ts
L
a
a
a
s:
FLLes ELR s'
t
DFNt49
MIAR:
ER
iD
MC3
ALsF
!a"
a
a
a
t4
I
2af
'
LAX
iFL
1
;Me;
Ni
"
am
a1
14
:3
anINC8: 364Al
a a
aA
ATL
: 39TLP4
SF026 LG A
L1
~
aU
a
1
ATUU
731
L1C0
:A i
ss
:I
E1R AT:
TL
FO
82
i
FLL
BDAUL D
~
~
s : AGA :
A
s
aL
2W'T7
w
oo
ON
CL : 6
CO
2
:
2
is
M
L 26
FLLIn
AL
5
i4F17~1F
I
97001 06166"
~aC
1
2~.
tsos 6
5
ORD:
sEl4R:
1A
s
LAX
~
MIA 3s
814
'
_
ATL
6j-.3?rfml
tis
25
2
19
SJU
:
_
110
sl
:
FRIDAY
FM
44 ATL PBI
LGA
LAX
0a
aqe
%6 ATL is
a
aa
:MCO :
84
ATL
:14
2eWEKL
LAX
as :
ATL
BOS
14
LAX
a-
a
LAX
:34
FLL
tt
:
s
ATL
i
tuATL
ATL
se
':
i:s
NIA: s7
si
ATL
ORD
ATL :78
_
_
SEA
ut
DFN:
s
LAX
:290 _1I8
n
: ENR
NCO 1AU Ali
lit:
BDA:
6S
:LAX
oz
1-10
SF0:
0
04
03
02
o
SATURDAY
SUNDAY
s
NCO
FN
:0s
:
ss
BOS
6s7
66:
:
__
NCO:
L~9100
ALGORITHM
Monday
THURSDAY
OF
as'
TPA
3
21
to
PBI
NCO 261
20
3M
i e
261 NCO
006010,
3
7igg
8
ATL iss
s
043;
003 pjg
1
SJU
SFO
SEA:
1
17
a
ATL
7
ogiga,
006 00100
QUIT
MONDAY
WEDNESDAY
delta-new
02
1-05
WINDOW
FILES
TUESDAY
ROTATION CHART
00
SEARCH
EDIT
REPORTS
OTHERS
I,
a.. a- a- a~
I
ATL:
is%
FK
ts
LGA:
BOS
sem:
ATL
46:
SF
21
n
FLL
150
FLL
ist
008001003
FIL
:364
ATL
id
04st
Lo'00100,1
DFN
a
1
ATLus
a NCO
ATL
244
:ATL: iss: PBI
14
FLL
:12s
ATL: @2: NCO
1m
q.;s349 1
a
SFO
lse
:
6Aa
87'
FLL
a
LGA
7s
ENR
DF
C
its
a
FLL
5s TA
6'3
10 0
3.
14s
0FLLu(1C0
a1a
a~~
IA
M443
1
31ENR
is 'ATL|
ENR
|14
aI
FLL
ENR s2s
aes
112I""=
144 Z
@"
2
BDL
OS6
:FK:
6e
|ATL
1%
|761
NCO
121
12'
2
FLL
ATL
6
2M
delta-new
ROTATION CHART
01
se
:
:
ER:
ENR
ORD: 7s'
ATI:
79'
AL:
ORD:
::
4
:
91
SJU
:36:
SJUJ
:s;
116 PBI:
:ATL
:ATL %6 PBI:
b
S..
i~
_W__ 04
- a - 02
:7-SJ6Jis2A222
__s' 4ll 15-12EHkRD
Eu MI,: R:
Tu pl1:
AGOIH
UT
WEEKA A
WNDOW'
S
SAC
EDT
REPORT
OmHERS
I
av
Th glg
detae
Th soe
:
Al
79'
ORD:
NL
s
Raa
D a
:
a
s~
a
aeekl
:
sJUi
ATI
:6:
sAT
J
11,1 ruT:
B
i
Fr glo
se
ENR:ALSU6
Sa
Su u
ATL
9:
Rs.:ENR:
ER
OD
Su~3:N
a
Tu
a
aD
a
a
SRD
a
T i PB
a
a
a
a
00
a
S'
ATL
s
ATL
a
He
ATL : ~SJ
so
a
IV if
All
OWs'
.
ATL:L
:ATl::
:ATL
ATL:
IVI
51
N
02
a
a
:LA
aB
a
ATL
ss2~
SJU329
a
Ai
96
:s6
CO:36:
AL s NCO
e se
: IIJ
to
NCO
ALliss'! NCOits
F0: :so ALliss'
SFo: ss S o As Ai TLissA MCOis6
D a0 ASF
TL
a
SJU
oti
if
ss~
DFN
109:1
:DFH
DFN
DF
s2
:LAX:
:.2
:LAX:
LAX
:LAX:
:2
:,2:
,7
ThI
Fr
=
SF0
se5
ATL:
_ 1
0
m
ts
AL
eDF
ATL
s
7t5CrL
Sa
As
SE
TupooI
ATL iss
NCOisi
0L
DFN
32:,
L
LAX
A
a
JE
4 13~i
OTHERS
-
ELI
110
U
fREPORTS
____
EDIT
~-...
___
SEARCH
WINDOW
Fit
ROTATIO N-.CHA RT
EEKDAY
QUIT
ALGORITHM
3CE
UEL.
EQ FROM
0_
DEE
ARgy
CLASS MEAL
EE
STATION CHA RT
STATION PROF ILE
L10
XG FLL
JFK
1720
1948
8511
62
L10
XG
BOA
BOS
1650
1757
851110-999999
014-014-014-014-014-000-014
62
L10
X6
BoS
BDL
1844
1925
851110-999999
014-014-014-014-014-000-014
63
L10
XG
BOL
BOS
0700
0744
851110-999999
021-021-021-021-021-000-021
63
L10
XG
BOS
ADA
0855
1155
851110-999999
007-007-007-007-007-000-007
65
L10
XG
ATL
MIA
1012
1150
851110-999999
008-008-008-008-008-000-008
66
L10
XG
LAX
ATL
0000
0534
851110-999999
009-009-009-009-009-000-009
66
L10
X6
ATL
sJu
0704
1105
851110-999999
006-006-006-006-006-000-006
67
L10
X6
ATL
FLL
1326
1505
851110-999999
023-023-023-023-023-000-023
70
L10
XG
ATL
ORD
1644
1735
851110-999999
006-006-006-006-006-000-006
74
L10
XG
P1I
ATL
1420
1600
851110-999999
012-012-012-012-012-000-012
74
L10
X6
ATL
BOS
1705
1920
851110-999999
007-007-007-007-007-000-007
76
L10
XG
FLL
LGA
2025
2300
851110-999999
011-011-011-011-011-000-011
77
L10
XG
LGA
FLL
1955
2240
851110-999999
010-010-010-010-010-000-010
79
767
X6
ORD
ATL
0815
1100
851110-999999
001-001-001-001-001-000-001
80
L10
XG
MCO
ATL
1105
1222
851110-999999
014-014-014-014-014-000-014
60
6UOSPRDFILC
s
015-0 15-015-015-000-015
.'
~
--
WINDOW
ROTATIONCHART
F
SEARCH
- 05
881001
-
ALGORITHM
-
CHART:
SSTATION
|
|Saturday
QUIT
EEKDAY
SCHEDULELISTING
----
TIME ZONE :
ATL
DATE
EDIT
REPORTS
4'OTHERS
STATION PROFILE
SUF PROFILES
LOCAL
01
00
02
03
04
05
07
06
08
10
09
11
12
13
15
14
16
17
18
19
21
20
22
24
23
iL
3:9
94
NIa
PI
176 112 1933 98
SF8
FLL iCU Ell IF
354
ilI
64
FLL
IC LAI
115 117
17
Ili
ili
449
P
799
Si8
Ell
194
MCI
195
SF6
97
Iit
31
07
08
09
10
11
12
13
14
15
16
17
32
El
116
SF8
96
SJi
146 177
LAI ORD
123
EC8
2619
Eli I
+
M4
IC
761
Lil
187
I
8
IC
1
155
06
111
NCO
1
L
83
sit
33
)5
16
LC
+
44
4'1 4-4,4SF3
137
97
134
LAX
158
NCO
32
SJi
36as
Ell
886
98
PI
116
It
96
TPA
156
ice
261
141
18
19
20
21
22
23
24
01
02
03
04
05
GMT
l\)
EDIT
FILES
EARCH
ADD
OTHERS
WINDOW
WEEKDAY
QUIT
ALGORITHM
COPY
ATL
DATE : 881001
b ELETE
- 05
TIME ZONE :
s:iLETALL
|Saturday|
INTEPCHANdik
MOV1E AU-L
MODIFY
LOCAL
01
SWAP
02
03
04
05
06
07
UNDE LETE
08
12
13
14
15
16
I
17
18
19
20
21
22
23
24
goo
1341
LAXI
j354
I[i
64
FIL
57
1Ii
41,
14"4, 1
151
IeCI
32
Sig 1
4
4]
4,
srI
I
MIb LAI Ell
PuIT BP
SF6
137
7
115 110
755
33
4
-
4,1 +
-41
445
194
135
33
57
06
GMT
07
08
09
10
11
12
13
14
15
16
17
761
17
Ei[O
go
156
116
6
L
155
123
Ied
26131
Eli IF
IF
cd
3
1
83
56
Sig
146 171
LAI Sib
-U, 4-+
AW?
4,
Sao9
05
16,
Lgi
tt 1 3
116
ieC EU SF3
-4,
44
I5
Ali
261
96
55
141
18
19
20
21
22
23
24
01
02
03
04
05
OTHERS
DATE : 881001
EDIT
SEARCH
FILES
WINDOW
WEEKDAY
QUIT
ALGORITHM
- 05
TIME ZONE :
ATL
REPORTS
|Saturday
|
LOCAL
01
02
03
04
05
06
07
354
Ili
J64
FLL
+
08
09
10
94
FBI
t76 112 73
FLL NEIII
[1
+
+
SF3
137
115 117
7
449
12
13
GIGice
14
15
$:s
IIe
83!
194
18
16
LGA
32
I 11 ECU
![1
~j
Il~~~~iijEI
57
ii*[
S0
SF8~~
SF3
97
Sig
13
83
I
5
19
20
IF
31
I
1
L
I
21
56
SIN
16
146 177
F
I
1
1
S9LI
IIi
%L11
B
to
799
195
17
EI
FBI
Egg
16
134
LAI
15
I&
11C
LAI
I
11
22
23
123
ECU
261!
l
s
Eli
98
335
336
FBI
116
oil
96
761
Lil
137
24
TPA
156
NE8
261
141
6i7
06
GMT
07
08
09
10
11
12
13
14
15
16
17 18
19
20
21
22
23
24
01
02
03
04
05
OTHERS
TIME ZONE
ATL
DATE
-
REPORTS
EDIT
SEARCH
FILES
WINDOW
WEEKDAY
QUIT
ALGORITHM
05
881001
|Saturday
03
05
|
LOCAL
00
01
02
04
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
134
334
oil
164
FLL
SF8
|'n
176 1U2
FLL NEI
7ts
EIt
83
IF
IC Lax Eve
I
137
7
Lax
158
i
Il:
Ia
.34
MC8
$F8
Po
115 117 194
799
195
83
Si8
its
Ila
449
97
so
8lt8
33
12
Si8
MIto
SF
88o
i3
06
07
08
09
10
11
12
13
14
15
16
17
SG
28
MCI
26L 1
Eve IF
Sig
146 177
LAI 8R8
IC
ElR
8s
886
PB
761
Lil
187
1
SiL
18
MC8
I
8
I
57
s8
TrP
116
156
MCI
EC
261
IN
1
1
56
141
iSE
)5
111
161'
LGA
12 116
E8
F8
19
20
21
22
23
24
01
02
03
04
05
HMT
OI
REPORTS
OTHERS
TIME ZONE
ATL
DATE : 881001
-
EDIT
SEARCH
FILES
II
QUIT
WINDOW
m
WEEKDAY
IIIIIIIIIIN
05
|Saturdag
|
LOCAL
134
LAX
158
lee
3?
Sig
J4L
SF6
97
lit
Ice
ILI
329
EVE
-4-41
116
SF3
146 96
tax Sig
4-l
LA:
117
24
ALGORITHM
116
OTHERS
EDIT SEARCH
REPORTS
FILES
WINDOW
1
WEEKDAY 1 QUIT
L10
394
L 10
MCO
0240
0434
FLL
0609
0030
0454
MIA
0634
0526
0030
U
112
110
ATL
0030
0746
MCO
0902
0030
194
L10
LGA
0030
0730
PBI
0910
0340
0030
0030
83
110
839
L10
98
ATL
0100
0749
DFW
1045
Oils
0915
MIA
1058
0052
SFO
1137
0030
0431
L10
702
SFO
0520
0605
ATL
137
L10
0604
MCO
0715
0030
ATL
197
L10
0639
MIA
0815
0100
ATL
449
D
L10
0932
LAM
1101
0139
ATL
117
0
110
1023
EWR
1215
0120
FLL
194
L10
1027
MCO
1140
0110
-DFW
195
L10
Ps1
1320
799
D
L10
1157
SJU
1600
0105
ATL
83
D
L10
0115
1200
ORD
1309
0427
ATL
83
D
L10
0030
1207
MCO
1320
0105
ATL
155
D
TIME ZONE
881001
D
L10
BOS
0035
1155
PBI
1422
D
L10
LAX
0112
1655
ATL
1865
:-
L10
05
0400
PBI
0700
0030
ATL
494
D
LO
1530
ATL
1750
0107
LAX
167
D
LO
0108
0108
158
0326
1150
0052
KI TES
-05
TIME ZONE :-
881001
364
ALGORITHM
LO
REPORTS
I[-- ]E*
I
STATION ACTIVITY
ROTAT ION CHAR
06
Monday
ROTATION SUMMARY
FLIGHT
7
R
I
AUDIT CHANGES
LID
SCHEDULE COST
61
:
us
DFH
MIN GROUND TIME VIOLATION
003
10
SIFa
L08
44A TL PBI
LGAs
LAX
AT137:CO:1IZATL
26
0U8W11
'
010
01
LIs
JFa
12
FLL
:364
ATL
44
FLL
MIA
a
L13 ool
FLL
t7s
4
LAX
as
56
TPA
L16808
====
um
ENR
a14
a85
a
a
aTm
_=
2W~2I
14
NCO
ATL: t2
BOS
se
n:FLL
2Al:
FLL
LGA
76
s26 : ER
DF
D13
se
NCO s ATL:
ATL is
aaATL
Lo15 g 108
s: PI
ATL
:
JFK is
1
g
14:
FLL
a2aa
EHR
6s
200 1.
:IDA:
62
BOSs1L
l
F
:6s:
EMR
:
32507;?
ATL is1
FLL
tt
a
is% :LGA:
FLL J44
at
lOS:
14
146:
134'
AL
sw
ORD
a
LAX
= 0: ILI
ATL 261
261
DFN
ATL
si
ssIA
ATL
:t$
ATL
is:
OA:
LI10
LI0
se
is
ATL
:2
SJU
SWITCH RESWITCH
ER
14i
LAX
'
SEA
LAX: n
MCO tit AU.
tui
BOS
676
IRREGULAR BLOCKTIME
261 NCO
: LAX
s2
::
DFH
1
a
DFN
its ATL: sl
ss5 NCO
EURDATL
LAYOVERS NEEDED
NCO J
s
MARKET SUMMARY
L02 3 glo
057
ATL: I
S1
3
SCHEDULE
aAE
LI0D
ALGORITHM
QUIT
WEEKDAY
WINDOW
FILES
SEARCH
UTILIZATION
JFK:
NCO
is@aNCO
lzt
1s2
FLL
ATL ts6
2
SEARCH
REPORTS
OTHERS
8
so
87
-e
Monday
delta-new
ROTATION CHART
05
e9
16
1
2___3
1,4
15is
ORD
ENR:
L10
17 i
10
ATL
:
|
|
UVaWe
89ITT3U
A
FLL
I
364 ATL :44
a
:FLL (1iMCO
14
SP
2
ATL:
14
PRI
pesmm
1m
ae40
is
FLL
ATL: 112 NCO
imm 1 woK;
SFO
O
ENR
126
1
134
ENR
:ENR:
005
t
1
msao
| FLL
6es
sammu
-
LGA
Ipew
too
Os FN
NCO
2
:BDA:
s
MCO se A:
|
FLL
10
isi
JFK
144
n
LGA:
is
04.0 iu
FLL
BOS
se
ATL
:46:
i34
gsm
m
FLL
t6
B6S
i
ATL
10
Lo6 ulgg
14
21
FLL :176 ATL us
TPA
fmm
1eewt
vi s :ATL iss:
1m
UiIP87
t4FM9
MIA
13-107~~~
1
:T
ORD
I
L12g 68168
L15008
15
s
TL
LAX
u
ist
JFK
I
AT
DFN: sw
71
1m
I
20
s:
:S53I
LoI
ATL 2s
n61
le,7
I
ATL
;
a00016
g
I;
is
s7
LAX
10
:
SEA
'1
ATL m37:MCO : 1ATL
%s
6MIA
es
:LAX
70
: ENR
i
LAX
ATL
BDA:
s1
BOS:
4S4 ATL PBI :1s4
LGA
LAX
t£1
e,4
PRI:
12
:LAX
15Is7a
as:
:2'
pee
NCO:
93
u6
7f5
pt
1A54
02
DFN
uS
1:-24
SJU
ss
ATL
el
ATL
ol
NCO tit ATL
:DFN:
1oepee
3
ATL:
73
MIA: 3s
NCOsk
24
:m
BOS in
TPA
105
MCO is
see5
,eye
23
: ss6
: DFN
676:
1
22
: SJU
:
1m
eis
PBI
0040@s NCO
21
32
3'2pi
:
056
s
2,0
t
:LAX
iie
1,9
ATL tss
43
SEA:
is
ATL:
1
so
SFO
007 ppgisameam
SFO:
2088
008
LOIH
QI
WEDY
WINDOW
ES
ROTATION
._.
....
..
GMT
UME
EDI' FLGH
62'
FLL
s
ENR :82'
BOSsa BL
JFK:
6s
ATL : i
t6W68
12
gam
NCO t2
|FLL
1 M:t "p.
ATL
I'S6
"PKA:
29g"
, , ,, 1"'1"
"Alfill
01
01N
hi
'Wil
1,
, I1111
111W
MINIMUM
11000MINNOW1114,
-221-
VI. STAGE 2 - AUTOMATED DECISION SUPPORT
INTRODUCTION OF AUTOMATED ALGORITHMS, EXPERT SYSTEMS
- to assist human schedulers with certain sub-problems
- to eliminate manual effort at certain steps of process
- to broaden search for optimal or good solutions to
scheduling sub-problems
- may introduce mainframe, large scale optimization
algorithms
EXAMPLES OF EXISTING AUTOMATED DECISION SUPPORT
ALGORITHMS
a) Best cancellation of flights given breakdowns and spares
b) Least revenue loss when reducing available fleet
c) Optimal switching of flights between types of aircraft
d) Automatic switching for transition to new schedule plan
e) Automatic weekend schedule cancellations
f) Automatic holiday period rescheduling
g) Minimum fleet size for given services with time windows
h) Automatic gate assignment at all stations
i) Automatic aircraft rotation generation (with maintenance
constraints)
-222-
VI. AUTOMATED DECISION SUPPORT
a)
b)
Best cancellation of flights given breakdowns and spares
Least revenue loss when reducing available fleet
Fleet Routing Models
- use network flow algorithms
OBJECTIVE:
Maximize Operating Income
GIVEN:
-
Set of potential services to be flown with fixed operating
times and known net operating income
Daily ownership costs of aircraft
Desired overnights
Fixed number of available aircraft
OUTPUT:
- "Best" services to be flown
- Marginal value for services not flown
- Marginal value of adding an aircraft to fleet
WEAKNESS:
- Fixed service times
- Fixed net income for services i.e.no spill if not flown
- Single type of aircraft-solved sequentially
AMIR
-223-
SCHEDULE MAP
STATION
STATION
(A
STATION
6
C0600
A0600
Time
Cycle or
Overnight
Arc, C
i j=
Ground
arc, G
lij
0
=
U- -=
Service Arc, S
I Jj
=
=
c-
=
0
A0559
c. .
'I
=
00
ci - = daily
0
rental cost
of aircraft
o
value of service
C0559
4I
i
LI L
ROTATION
12
eMT In
LAX:
a
a
a
a
a
delta_jvctst6
19
1
17
6
CHART
is
a
a
a
a
a
14
a
a
a
a
a
15
ORD:
6~1
tre
ses~ ; FII
ORD
e?56
DFN
inz
SF0
a
a
a
a
a
a
a
a
a
a
a
a
a a
a
g
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a a
aa
a
aa
a
a
a
a
a
a
i
a
a
002 00100
a
0-0100
a
a
a
a
a
012 818
013008
01500100
g
a
a
a
a
a
a
a
a
02
03
ORD
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
1~7
j~~a
a
a
e4
0
U5
,
L~X
AlL : ss
ORD
t72
SF0:
22j2~1/
a
g
a
01
22J231
a
a
24
RDa
1
a
a
23
:ORDU
'i;igu
006
016
0--011881
00800100
22
ssa
:i
g
a
007
21
A TL
a
005
ALGORITHM
QUIT
Monday
20
:
:si
WEEKDAY
WINDOW
FILES
SEARCH
EDIT
REPORTS
OTHERS
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
-
-
CMTN
IV
)
a
1I
12
al
1n
I9
7_10
L
20
21
22
L00B
ORD:
s~
LAX:
DFN
i
SF0
92
al
:OR[D
I
II
II
II
7i
II
*I
II
*I
*I
007ole
*
I
I
I
I
I
I
408
I
*
00901oo
I
007 00100j
0080100
I
I
I
I
I
I I
I
I
I
I
013011
I
I
I
*
012011
I
I
I
I
I
*
0
I
I
I
I
I
I
I
I
I
I
I
I
~'~
I
I
I
I
I
I
I
g
I
I
I
I
I
I
I
I
I
I
:ss
I
I
I
I
I
I
I
I
I
I
I
Ig
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
*I
I
I
I
I
*I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
014661001
I
I
I
I
016ggi00
as
I
I
I
I
I
I
I
I
I
ORD
8
8
07
I
I
I
I
I
I
I
I
I
*
I
I
*
*
I
I
I
I
i7
-7217
,
I
I
I
I
I
*
I
I
I
I
I
I
I
I
I
*
I
I
I
I
I
-
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
*
I
*
-
I
I
I
I
*I
*I
*I
I
I
I
I
I
I
I
I
II
I
II
II
II
I
*I
II
95
I
*
I
g
I
I
I
g
Trjq
11
1
I
I
I
I
I
*
*
94
*
*I
*
I
I
I
I
I
I
I
I
I
ATL
*I
II
oI
*I
*I
II
16
Ga
: ei LAX
OR
:1s:?
:sn
24
ORI
:.::ORD:ses:DFH aim :ATL: se
00
23
I
I
I
I
I
I
I
I
I
I
I
I
I
*
*
*
I
I
I
I
I
I
I
I
I
*
I
I
I
I
*
I
*
*
*
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
*
I
*
I
I
I
*
I
*
I
*
I
I
*
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
*
I
*
*
I
I
*
*
I
I
I
I
*
I
*
I
*
*
I
I
I
I
*
I
*
*
I
I
I
*
I
I
I
*
I
*
*
*
I
I
I
I
I
I
*
I
I
I
I
I
I
I
I
I
I
*
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
S
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
-226-
VII. SUMMARY - STATE OF THE ART IN COMPUTERIZED SCHEDULING
Conclusions
1. We cannot create analytical models which are adequate
to describe mathematically the complete airline
scheduling problem.
2. For existing models which promise utility, we generally do
not have the correct data inputs, and it isdifficult to
conceive of creating the necessary models for
passenger behavior in today's competitive markets. The
existence of large scale solution techniques is not
sufficient to justify their use at present.
3. We can provide quick, accurate answers to many subproblems which occur in the complete scheduling
process, but we need an environment which allows these
techniques to be available to human schedulers. This
environment isnow available in the form of a network of
computer workstations.
4. It isattractive to consider a single, integrated system to
be used by various airline personnel as the scheduling
process moves from initial planning to final execution.
5. People will remain an important part of the airline
scheduling process. They are responsible for generating
good schedules, and need "decision support" in their
activities. There never will be a "push - button" scheduling
system.
6. The desired approach isan incremental introduction of
computerized assistance via graphic workstations. The
strategy should be to create evolutionary stages:
Stage 1 - Introduce the Scheduling Workstations
Stage 2 - Introduce Automated Decision Support
Stage 3 - Extend to real time Execution Rescheduling
AN,
I
111HINIIIIII
III
III ii
I,
INN
-227-
VI. SUMMARY - STATE OF THE ART IN COMPUTERIZED SCHEDULING
(con't)
7. The scheduling process is not permanent
- as time goes by, the problems change (perhaps
temporarily), and the markets evolve, and there will
be emphasis on different aspects. It will not be
possible to create a completely automated
decision maker which keeps up with changes.
8. As these tools are developed, they have their impact
on the Scheduling Process
- it will change in its flow of information, the sequence of
processing will change, and eventually the airline's
organizational structures will change. The
introduction of computer automation must be
adaptive to allow these changes to occur.
9. Every airline will have to develop its own automated
scheduling system and manage the evolutionary
impact on its operations. There isno single, turnkey
solution to be provided by outsiders. A conceptual,
long term plan is needed to direct the evolutionary
effort and prevent building an incoherent set of subsystems.