FTL REPORT R80-3
PILOT WORKLOAD IN THE AIR TRANSPORT
ENVIRONMENT:
MEASUREMENT,
THEORY, AND THE
INFLUENCE OF AIR TRAFFIC CONTROL
Jeffrey G. Katz
May 1980
FTL REPORT R80-3
PILOT WORKLOAD INTHE AIR TRANSPORT ENVIRONMENT:
MEASUREMENT, THEORY, AND THE INFLUENCE OF
AIR TRAFFIC CONTROL
JEFFREY G. KATZ
Flight Transportation Laboratory
Massachusetts Institute of Technology
Cambridge, Massachusetts 02139
May 1980
ABSTRACT
The operating environment of an air transport crew is characterized by multiple interrupting tasks, these tasks being
composed of a mixture of purely physical control and purely
Measurement of crew workload is
mental planning processes.
thus d difficult undertaking due to the necessity to resolve
workload contributions imposed by several
sources include physical efforts, mental
task interruptions,
tribute,
subjective
sources.
efforts,
and emotional disturbances.
opinion rating scale is
usa as an effective measure for this
environment.
These
random
A multiat-
presented for
air transport cockpit
An analysis is performed which indicates tnat a major component of workload is induced by the federal air traffic control system.
Mechanizations of this loading inClude speed
and altitude restrictions imposed by regulation, confinement
and restraint imposed by the structure of the National Airspace System, and loads induced by a stochastic interruption
process associated with ATC voice communications, In fact,
the analysis of a routine transport arrival into Boston's
Lagan airport indicates that
the (primarily system induced)
workload levels in the terminal area, may be higher than the
(primarily aircraft
induced)
workload levels on final
approach.
A fixed base,
3oeing 707 simulator was employed to investigate the consistency, sensitivity,
and acceptability- of the
31,
subjective rating scale.
Four airline pilots and four
general aviation,
IFE pilots flew a series of routine,
IF1i
arrivals from hign altitude cruise into Boston's Logan airport,
each arrival terminating with a
standard instrument
a-proacit.
Consistent ratings were achieved across the airline suojects for all segments of the arrivals.
In general,
all subjects
metaidulogy.
seemed receptive to the
subjective assessment
:ABLE CF CONENTS
page
Chapter
I.*
INTTFCDUCTICN
:otivation
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
I
.
CONCEP:S
OF
Workload
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
. .
.
.
.
2
.
. .
.
.
.
.
4
.
.
.
.
4
.
.
.
.
.
.
11
17
.
.
.
.
22
.
...
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
Workload
Mental
and
.
.
THECRY .
WCFKLOAD
.
.
Objectives and OrganizAtion
.
.
.
.
.
.
.
.
Functional Definiticns of Mental Workload
Workloaid Capacity and Related Theory . . .
Surrary
I1.
V.ZESUZES
.
.
.
.
.
JF MENTAL
Physiological
Brain
upil
Heart
.
.
WORKLOAD.
.
.
Dilation
Eate
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
leasures of Mental Workload
.
.
. .
:ask Measures . .
Primary
Mental Capacity
Opinions
Sutjective
.
.
.
Variability
Spare
IV.
.
Measures . . .
Electrical Activity
Dehavioral
Suamary
.
.
.
.
.
.
.
.
.
.
. .
. .
.
.
.
.
.
.
.
.
.
.
33
36
.
.
.
.40
.
.
.
.
48
50
TFAFFIC
.
.
.
.
.
.
.
.
.
.
.
33
.
.
.
.
.
.
.
31
.
.
The Airspace Loading riechanisn
:egulatory Loading iechanisms
3echanisms
29
.
.
Loading
.
.
.
.
.
.
.
.
.
.
.
25
26
.
.
. .
.
.
.
. .
.
24
.
.
.
.
1
.
.
.
PILOT WOIKLCAD AND THE IN'FLUZNCE CF AI
CCNTE.OL
1
.
.
.
.
.
.
.
.
.
.
.
.
51
.
.
.
.
.
.
.
.
.
.
.
52
53
.
.73
.
75
.
.
75
.
Discussion ot the Reyulatory and Airspace
.
Mechanism
Ac
V.
*
.
*
*
* .
.
Expiicit
Loading
.
Implicit
Loading
.
WOr.KLOAD ANALYSIS
.
.
OF A DCSTON
.
. .
:he Scenario
.
Analysis Tools .
.*
. *
.
. .
*
.
.
. *
Cockpit Activity Tirelines
.
Task Precedence Maps
.
.
i
.
.
.
.
.
. .
.
. .
.
77
.
.
.
.
91
*
*
,
*
.
.
.
.
.
.
.
.
92
94
.
.
.
.
95
.
97
.
.
.
.
.
.
.
.
*
.
.
.
.
.
.
. .
.
.
.
*
*
.
-w
.
. .
AERIVAL
*
.
.
.
.
.
Voice Communication Mechanisms
.
. .
.
Analysis of ATC and Intercrew
Communications
An
.
.
Analysis
Busvness
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
.
102
.
.
.
.
.
.
.
103
.
.
.
.
.
.
.
104
117
122
..
.
.
.
.
Mu ltiple Tasks and Interruptions..
Simultaneity
Priority
and
:ask
Tas-
Constraints and
.
Slack
.
.
Iriggers
ATC and Intercrew
Concluding Discussion
VI.
.
.
.
.
. .
.
.
.
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
129
.
.
.
.
.
.
142
Communicatiar.s
.
.
.
.
.
133
FLIGHT SIAULATOR EVALUATICN OF A SU3JECTIVE RATING
SCALE
.
.
.
.
.
.
.
.
. .
Experimental Ctjectives
Simulation Facility
Flight
Experimental Scenarios .
Experimental Procedure .
Preflight Briefing
.
.
.
.
.
.
.
.
.
.
...
.
.
.
.
163
161
162
165
.
.
.
.
.
.
.
.
.
.
Concluding
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
146
147
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
146
.
.
.
.
.
.
.
.
EIB3LICGRAPHY
.
.
.
.
Warmup Flights
Data Flights
.
. .
Discussion of Results
Discussicn
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
150
157
.
.
.
182
185
.
page
Appendix
A.
SCE
a,
FIGUE;S FOS THE ANALYSIS
C.
EX?IENCE
COLS
F70i
WCRKLOAD
ANALYSIS
.
.
.
.
.
.
OF A BCSTON AFEIVAL
QUESTICNNAIFE
I ii
.
.
..
.
.
.
.
.
.
.
.
189
,
.
210
.
.
237
.
A
Chapter I
INTEODUCTION
Pilot worklodd has long been
and manufacturers
a concern tor the operators
'his
of aircraft.
stemmed from
some threshold of workload,
theory which states that beyond
a pilot's performance will be
concern
degraded to unsafe levels and
his ability to cope with unexpected emergencies will be seriously impaired.
Unacceptable levels of workload,
jeopardize safety
in a
system where
the costs
therefore
associated
with error are intolerably high.
L3APT 0rL4
1.1
The late
that brouqh t
1970s saw a
increased public
Siirilarly,
safety.
series of major
attention to
757.
recent,
the Douglas DC9-80,
These events
issues of
air
the early 1980s will witness the intro-
duction of at least three new technology,
aircraft,
airline accidents
are
all
transport category
the Boeing 767,
significant
and the Boeing
contributors
to
heightaned research efforts in areas related to the
human factors of cockpit design and operating safety.
at this time of increased
public concern,
For,
there remain few
acceptable answers to the following questions:
1.
Should an
aircraft have
a two
or
three man crew?
2.
Against what
standards snould
safe
new cockpit
and
use of
instrumentation
be
the
displays
certifi-
cated.
Against what
3.
standards should
new
flight procedures be evaluated.
Tahe answers
to the above
understand
crew.
and measure
questions require the
the wor&load
of
ability to
an air
transport
At present neither can be done to the satisfaction of
the aviation community.
CgBJECTVES
1.2
"Nj CLGANZATION
This thesis research was conducted as p4rt of a multidisciplinary
effort
1978-1933.
It
at
MIt
over
the
addresses several
it
two
academic
aspects of
the workload
addresses the state of recent workload
problem.
First,
theory as
applied to the duties
and environment of
transport crew operating under present-day Instrument
Eules
(.:Fa).
measures
these
Related to this
is discussed
:neasures are
erivironment.
years,
with a
adapted for
the literature
particular
use in
an air
Fligtt
on workload
emphasis on
the air
how
transport
Next,
the linkage
traffic
these
or interface between cockpit
control systemt is
two systems*
other.
A
strongly
The activity in each of
influences
framework for analyzing
cribed that,
nificant
examined.
in
particular,
influence of
'is
activity in
this interface
useful in
traffic
the air
and air
the
is des-
examining the sigcontrol system
on
pilot workload.
Finally, a complete analysis of a Boeing 707 arrival into
Boston is described.
This scenario serves as
baseline for a series of
were devised
flight simulator experiments which
to evaluate
developed at :"T.
a pilot
workload 'reting
These experiments
explore some of the more
the initial
an Anglytic
scale'
also were designed to
theoretical constructs examined in
portions of this
work,
The
flight simulator
experiments and their results are discussed.
Pilot worklodd is,
however,
very complicated
stand conceptually.
is
today,
a 'topical' subject.
subject to analyze or
It is,
to under-
The fundamental objective of this work
to place pilot workload in'perspective by identifying and
examining key aspects of the problem.
Linatj.on of theory,
In this thesis a con-
analysis and experiment are employed in
this task.
*The aircratt systems includes crew, aircraft, and aircraft
lynamics. The AIC system includes system architecture, the
controller, ani other aircraft.
-3
-
Chaptor 1I
CONCiPTS CF WORKLOAD THEORY
nis chapter lays
a framework which serves
discussions of
for the
as the kasi-s
The concepts
later chapters.
of
are elaborated by collat-
workload and workload measurement
inq material from existing literature and troa research perthe present
conjunction with
formed in
thesis.
this
In
chapter the theory and conceptual description of workload is
explained in a
techniques for
workload measurement
casses
The following
general sense,
chapter disair
the: IFE,
transport application.
2.1
WOEKLCAD U.D MENTAL 1_CAELADA
Historically
the major
study of
aperator, workload
Tha t is,
4ancentrated on physical workload.
was
researchers and
designers were mostly concerned with anatomical restrictions
and liMits of human capabilites in terms of strength, effort
and physical exertion.
mental,
became apparent that purely
non-physical effort could, also seriously impair an
operator's
quently
Later it
his potential
performance or
the literature
has
more
impoutance of mental workload.
-* 4 -
to err.
Conse-
recently recognized
the
his
is
It
workload.
or mental
total worxload
either
exclusively with
almost
deal
thesis will
convenient
to
state that
Total Workload = ?hysical W9rkload + lental Workload
Physical
workload is
and research
continues
the human operator.
Physical
quantifiable
today into physical loads on
terms of hard ghysiological- par-
in
workloa, d an be measured
perspiration rate
as heart rate,
ameters such
and carbon
dioxide expiration rate.
However,
in today's air transport environment the mental
workload corponterit
is
increasingly the most impor-
becoming
tant component of total workload
alone is
discussed in this
indreasing in
hias been
and,
thesis.
for this reason,
Today's
simplicity and automation.
transformed into that
The
ot a 'flight
pilot manages
it
cockpits are
pilot's job
manager'
his complex
(see
system
Lavcson,,1978).
As the
through the air,
indeed physical levels of effort are quite
be reduced.*
low and continually seemed to
cal
loads cannot
be
ignored,
Althougn physi-
inentil loads
have
become
increasingly more complex and crucial to understanding jilot
*See Hay, et al,
1978,
for an example
ceductiLons in the DC9 series aircraft.
.-
5-
of physical task
wor kloa d.
:o this end,
load is not
must be clearly stated that mental work--
it
directly proportional to
performance nor is it
the number of tasks being performed.
'intervening'
variable (see Simpson
is not directly observable.
mental workload is
mental workload is
and Sheridin,1978),
it
In generil it can be said that
some combination of mental effort,
mation processing,
an
and emotion in
infor-
response to task demand.
Mental workload has to do with a sense of mental effort, how
hard one
(1978)
feels one
As Simpson
is working.
an4- Sheridan
point out 'one person can claim t3 have a feeling of
great mental effort,
no mental
while another can claim to be exerting
effort at all
and
both be performing
at the
same level.
Sheridan and Stassen
the
human is
linkages
(1978)
total workloal
performance.
Figure 2.1 is
which shows,
in system form,
take place
tas k.
his system
interfaced to
through which
when a human
show in schematic
a
mince
and postulate
can influence
the
task
reproducti..n of this concept
the various interactions that
controller (operator)
works
on a
The figure theorizes how mental worKload is involved
in affecting task performance and it
ure,
Xorm how
that it
is clear, from the fig-
need not be directly related
The figure also shows
to task perfor-
external variables which may
be measwured (611,AM2 and 33) to be representative of workload.
It
should be noted,
thouyh,
that Sheridan's model of Work-
load and Pertorinance shows tae I
rect
variables to be only indi-
iteasures of workload.
For a
load it
theoretical and conceptual uaderstanding
is
necessary to have a frdmework tor defining mnental
workl.oad or a paradigm for
(1978)
have constructed
case which is
Sheridan arid Simpson
analysis.
a paradig
reproduced in
the piluting task,
affect
of work-
for the
figure 2.2.
pilot workload
They note that,
there are three distinct functions which
workload and which also affect performance.
1.
judgement function that
'here is a
is
dependent on
pilot
experience
and that will contain his long term
planning efforts.
2.
There
is
a
supervisory
which allocates time
be
spent
short
function
and effort to
on tasks.
real-time,
is
This
term
a
planning
function.
3.
in
There is the
function that
sory motor
in-bred task
dmnost
more fundamental sen-
skill and
reflexive
operator.
-
7 -
action
contains
re.presents
of
the
Assigned criterion workioad
GIVEN SYSTEM AND FORCING FUNCTION
HUMAN CONTROLLER
Assigned task workload
SYSTEMDISPLAYS
SENSORS
NROSEFFECTORS
''"""""OSLY
II Asindklkwrla
,.4...,MANIPULATORSPRCS
y)
(~~~>T~isturb\
CONTROLLED
isturb#W
Pnce
I
In forma tion
are
"m ntal
tional1
Emo
processin
workload
woworkload
Snfrbje
tiiona
nry
4SEODR
PEyiolgyclScnayPi~t
D3ananysDS
a
D3
d D.
Figure 2.1:
a --.
Paradigm
Control
Human Operator ---
Muni
wrla?
EnergyS.CONDARY
statef
toerrnea
(wrla?
THE PHYSICAL ENVIRONMENT
OERATOR
T4EHUMAN
JUDCYXrNT F'NCTION
HENVIRONMENTAL
FACTORS
& memory
Long term planning
1f
Ci
Experience
Consequences
_
Evaluation
f
I
Pi
g
-
Criteria for
decision makin
Task Arrivals
(Previewed)
y
SUPERVISORY
!f ~
FUNCTION
Performance
Monitoring
Short term planning & memory
Real
time, offline
Task effort allocation
Selection of
next task
1. Observe all tasks
2. Estimate Pj(Mi) given P
3. Schedul'e tasks
U- 4. Determine Mi
M
Task specific
s
Supervisory
Mental
Work
mental work
SENSORY-MOTOR FUNCTIONtS
AIRCRAFT SYSTEMS
Task skills
Real time,
on-line
n
i
B1
earning
Mi
B
Behavioral
actions
applied to
tasks
M
T
Figure 2.2:
Qualitative Paradigm for Pilot Mental Workload
Task
Competions
Pi, Task Performe
ihe
paradigm
characteristics
of
figure
of pilot
2.2 fits
motor
functions
will
require
is simple and requires little
skill level will expend
the
known
performance.
at sone s&ill levels,
Thus some pilots will
effort.
of
and pilot
workload
The paradigifr also suggests that
sory
many
virtually
no
senmental
feel flying an ILS approach
Yet pilots on a low
ef.iort.
considerable effort controlling tne
aircraft and will not be able to adequately perform the long
term
planning
juately.
or
short term
supervisory
For this less skilled pilot,
functions
then,
ade-
the workload
may well be very high.
This formulation
of mental
worikload also
the long *erm planning function can
suggests that
load the pilot,
as can
the supervisory function, even though sensory motor tasks do
not.
This
is
a
automation relieves
matter of
great practical
the pilots
of many
importance as
physical,
sensory
motor tasks (see foerger,1978).
Zinally,
it should be noted
framework theorizes that
that Sheridan and Simpson's
both the aircraft systems
operating environment affect the mental
pilot.
warKload of the IF3
The huindn factors design of aircraft systems reflect
the importance of this influence.
the operating environment is
and has
and the
However,
difficult to predict,
been generally ignored
in workload
inf luence of the operating environment,
tra.ffic control system is
the influence of
a priori,
studies.
specifically the air
the subject of Chapter 4.
-
10
-
The
2.2
FUNNTICNAL
D"F:NIIgjiS OF MEAL
WOE KLOAD
no yenerally accepted definition of mental work-
There is
laad despite yeirs of research iri mental workload and opera-
al,
of
-1978)
Force study
A recent FAA Task
tar workload.
cites
report
et
reviewed some common
the state of the art,
:his
definitiori,
(see Hay,
two
different
quite
appruaches to a definition:
aggregate of
the system ...
determined
is
'Level of pilot workload ...
by ...
placed on the
the task demands
the
pilot by
coupled with the actions required of the
pilot to sitisfy those demands.'
'Level
or
the sum
workload equals
of somatic
energy
expended and task difficulty multiplied by the duration
of this activity.'
In
functiotial deiinitions described
this thesis the
chapter 2.1 will be used.
in
the
present research
to go
"t is beyond the scope of
further than
this in
in
atteimpt to coin a new definition for workload or mental
to discuss in more
however,
workload.
It is useful,
detail the
theoretical form of the
mental
ween task performance,
ioad.
relationships bet-
effort and mental work-
The ensuing discussion draws heavily on the pre-
viodsly cited work of Simpson and Sheridan
w-he
interested reader
is referred
-
11
-
to
(1978)*.
Chapter 2.3.2
.
of
Lodej
:ormulation 11 a Workload
The IFF
piloting environment is characterized
system where the pilot is
tasks.
tasks must be
:his is essentidlly the
in
continuously working on many
The tasks occur randomly (at time tfor
and often
his thesis
be worked
and ior each
In
tas&,
i,
by which time the task must
td(i),
This scenario is
piloting task and it
tative of the
studied
reguiced a different amount
or effort to complete
have been completed.
on simultaneously.
control behavior.
on supervisory
there %as a deadline,
task i)
paradigm Tuiga (1978)
Tuld's paradigm each task
of time
as a
is
guite represenshown schemati-
cally ky figure 2.3.
In general,
on
task i,
it is assumed that task performance,
is a
function of
expended on that task6
model,
is
the behavioral
Note that performance,
postulated to be
a function not
effort, but of behavioral effort.
places some utility on a
task i
leads him to
errort.
=(P
Desired
terms
Simpson drid SLeridan (1978).
-
in this
of mental
The individual pilot
of behavioral
P(B=
effort
performance level aChieved on
and a desired utility
required amount
P.,
12
-
In
expend the
algebraic
M(ental (ffort
Z.by
M.
PI ied
p
pilot
Task i
toi
arrival
-
--
time
_
td i
_~iti~~h~
N*
-
iji@Thi1ThTh1Th]iillIillflfliillThliJIIll
-
'1
deadline
1
Other
nown
tasks
I
I
I
--I----1
I anticipated task I
I
I
I
TI14E
Figure 2.3:
Task Arrivals, Dgadlines
and Mental Effort
-
13
-
t,
present time
Task
Per formance
P.
P.--
inum Perfoinance
in
B.
i
Behavioral
.
mnif
Activity
B.
I
(A)
GENERAL FUNCTION
Task
performance,
P.
Behavioral Activity,
(B) STEP FUNCTION
Figure 2.4:
Task Performance Function, P (8
)
1
-
14
-
Zo meet
his own
performance criteria
expend the required behavioral
the pilot
effort.
will
This is shown
by figure 2.4.
It
is
postulated that there is a monotonic relation-
ship between behavioral acti~vity
figure 2.5
shows,
increasing
and mental work.
still
levels
As
allow the
same amount of behavioral activity to be performed with
lesser expenditures of mental effort.
Mental effort then is
neous
pilot.
workload
Mental
a time dependent but instantaP
parameter (5t))
for an individual
work is
the time
integral of
mental
effort
This work is
Mental Work =Jt(t)dt
0 '
dependent on both the task,
i,
and the
pilot, p.
The pilot works in a multiple task environment.
some time interval over which workload
is
measured,
In
the
pilot may work on tasks simultaneously, or nearly so*.
The total mental effort level,
summed over all
tasks,
and using a moving average of interval T gives
t .
t.p
p
XL(t) = 1/T*gJM.(t)dt +J M(t)dt]
i
t-T s
*It is a matter of contention
whet'her pilots work on tasks
simultaneously or whether they time-share. This point does
not affect the present argument but may affect the choice
of a model for a human operatgr.
-
15
-
Task
Behavioral
Activity B
levels of skill
Mental Work, M.
Figure 2.5:. Behavioral Activity as a Function of
Mental Work
, instantaneous
ment al effort
Mental Work
Rites
Time,
Figure 2.6:
Mental Effort Level and Instanteous
Mental Effort
-MIA -
t
As shown
in
figure
2.6 this
moving average
smooth the instantaneous work
neous work rate may exceed
tends to
effort so that instanta-
the average effort level at
times.
WORKLOAD CAPACITY AND RELATED THEORY
2.3
With these description of the concepts of workload theory
it is now possible to take
relates
workload to
one additional step.
a workload
limit or
This step
capacity and
it
relates workload to performance*.
Simpson
and Sheridan
state that
associated with each task,
tion.
Total mental
tasks and all
and
work is
there
is
mental
work
with each psychomotor func-
then the
summation across all
functions.
MP= E N.
Total Mental Work
ij I
i - task
Pilot p
j
Associated
with
-
function
each psychomotor
work level or capacity,
element is
some 'maximum
,
for a given pilot, p.' An
J
index of the utilization or loading of each psychomotor ele-
MCAP
ment is represented by
M1/MCAP
p =
*This section also
op cit.
p<p<i
draws heavily on Simpson
-
17
-
and Sheridan,
Saturation of this element occurs when P= 1.
It
is
ize the
then theorized that an individual pilot will maximutility derived
from a
certain performance
less the disutility of the mental workload.
level
dathematically
this can be stated as
p
pp
p p
Sax(U) = Max(; U()
-U
)
p
p
ML< 1CAP. for each pshychomotor element j
subject to
and P.> P.for and task i (min.
i-
Implica tio n
The
thejodel
most important
cussed workload
implication of
formulation to this
presence of multiple,
suggests
performance level)
min
the previously
thesis relates
randomly occurring tasks.
disto the
The model
(see figure 2.7) that at different times different
psychomotor elements are constraining
Different types of task load
unique manner.
the pilot's workload.
affect different elements in a
Consequently, there will exist tradeoffs in
performance on varying tasks that
objective function
cited in
occur when maximizing the
the formulation
task tradeoff occurs explicitly due
simultaneity of tasks in the
above.
This
to the multiplicity and
piloting environment
(see fig-
ure 2.8) In fact, queueing theory would suggest (see Senders
and Posner, 1976) that performance will degrade as the probability
of tasks
of higher
worked on, increases.
priority
than presently
being
This is shown by figure 2.9, that is
also taken from Simpson and Sheridan.
-
18 -
-Motor component,
Pilot
Mental
Work on
Task 1
j=21
Sensory component,
-
Motor component,
j=22
Pilot Mental- Work on Task 2
Figure 2.7:
Mental Work Capacities for Two
Simultaneous Tasks
Pilot
Performanc
on Task 1,
P1
Pilot Performance
on Task 2, P
2
Figure 2.8:
Performance Level Tradeoff for
Two Simultaneous Tasks
Overall
Task
Performance
1.0
Overall Workload Ratio
Figure 2.9:
Overall Task Performance vs.
Workload Ratio
-
20
-
identified three categories of
Simpson and Sheridan have
task in the IFR piloting environment that define a hierarchy
This hierarchy recognizes
of task priority and precedence.
the existence of a task multiplicity and interruption dimenThe task categories are
sion to pilot workload.
1.
-
Operating tasks
directly with the
aircraft
which
concerned
tasks
operation of the
be
must
handled
Examples include air-
immediately.
craft maneuvering and control,
and
auditing radio communications.
2.
tasks -
Monitoring
tasxs of criti-
cal importance that
may be delayed
for a
of time
short period
operating
formed.
tasks
being
per-
include systems
Examples
monitoring,
scan,
are
while
out-the-window
and instrument
visual
cross check-
ing.
3.
- tasks
Planning tasks
into idle periods
deferrable
of time.
Exam-
ples include retrieving ATIS information,
planning approaches,
puting landing data.
-
21
-
com-
monitoring task
If a
the
being performed,
the monitoring
deferred until
planning task will be
task has
both planning
with which
The rate and randomness
and monitoring tasks.
the
Similarly,
been completed.
operating task will preempt
occurrence of an
task is
a planning
is cued while
occur will influence the work-
the three categories of task
load level perceived by a crewmember.
SUMMARY
2,4
In the air transport environment the mental workload component is increasingly becoming the most significant portion
workload has no
thought of
mental
Despite this importance,
of total pilot workload.
It can be
generally accepted definition.
as some weighted
combination of
mental effort,
information processing, and emotion in response to an operator's task demand.
Theory would suggest that mental load is
not a simple function of task performance.
The task load
by a
of an air transport
randomly occurring series
tasks
tend to
another.
crew is characterized
of multiple
occur simultaneously
and
tasks.
These
to interrupt
one
Pilots are thus trained to prioritize these tasks,
ranking each task with a certain relative importance.
Cer-
tain types of task are
easily deferrable for sustained per-
iods (planning tasks).
Others are deferrable for only brief
periods
while those of highest priority
(monitoring tasks),
(operating tasks) must be performed promptly.
-
22 -
It
is
posta-
lated
perceived
workload of
the magnitude
of this
that the
influenced by
the
pilot will
be
interruption/queueing
process.
These simple
concepts of
workload theory
are a
basis on which to build experiments and analysis.
this chapter
digm of the
IFR pilot presented in
the further
theoretical discussions of
analysis of Chapter 5,
useful
The parais
Chapter 4,
used in
in the
in the experiments presented in
and
Chapter 6.
-
23
-
Chapter III
OF MENTAL WORKLOAD
MEASURES
is the sum of his physical
The total workload of a pilot
and mental workload.
the air transport environment,
In
the insignificant component of
sical workloads are becoming
total workload as cockpits transition
1978).
more
duties in
Indeed the pilot's
of a
to a state of automa-
simplistic designs
integrated,
tion and
phy-
et al
the Late 1970s make him
and supervisor
flight manager
(see Hay,
than the
'stick
jockey' of an earlier era or aviation (see Babcock,1976).
These include
the indices of
perspiration rate.
of methods.
measurable by a number
Physical workload is
respiration rate,
heart and
however,
is a rela-
Mental workload,
tively new concern to the engineer and operator.
cal loads have reduced in
of automated systems,
due to new
ing.
As physi-
magnitude due to the introduction
mental load may have
complexities and uncertainties
in
been increased
decision mak-
the most significant compo-
Mental workload is today
nent of total workload in many operational circumstances and
it is a concept that is, as yet, not well understood.
focused on measures of men-
In this chapter attention is
tal
workload.
measures for
Candidate
-
24
-
the air
transport
and evaluated
are discussed
environment
in
terms of
the
results of previous research and with respect to the conceptual theory outlined
An attempt
in Chapter 2.
made to
is
show how suitable these candidate techniques are for measuring the workload of an IFE, air transport pilot.
This chapter is not meant to
sible
measures
of
be a compendium of all pos(see
workload
surveys
measures
promising results in
have
seen very
which
the air
are
Wier-
Rather, this chap-
wille,1979;Williges and Wierwille,1979).
ter
instead
most
likely
to
yield
transport application or which
common usage
in
measuring aircrew
mental
workload.
3.1
PHYSIOLOGICAL MEASUEES
Physiological measures of mental workload are those meth-
odologies which seek to infer
nals are heart rate,
This
skin temperature,
cldss of measure
from workload
is promising
measurement by tapping
physiological signals
These, types of measures are
but there
is
because it
physiological functions of men-
and matching these to known
tal workload.
and brain electrical
the uncertainty and subjectivity
portends to remove much of
understood
Examples of these sig-
physiological signals.
quantifiable,
activity.
workload levels by monitoring
a great
deal
underway that should yield improvements
ological workload measures.
-
25
-
still not well
of basic
research
in the use of physi-
3.1.1
Brain Electrical Activity
These techniques utilize electroencephelograph
brain electrical activity.
of the
brain emit an
is theorized that the neurons
It
electrical stimulus
Belated Potential or EBP)
records of
(called
an Event
corresponds to the informa-
that
tion processing associated with a discrete external event.
related stimulus (also referred
The signal from an event
to in the literature as an evoked response)
is
imbedded in
Signal averaging
the
noise of an EEG
record.
low level and
techniques must thus be used to'
extract the ERP pulse which
typically has a magnitude of 10
to 250 microvolts and pulse
widths of 500 milliseconds
Figure 3.1 shows
using signal
(see Donchin,1978).
a typical ERP trace
averaging techniques
that was extracted
The signal
peaks are
labelled by their positivity and by the time after the triggering event at which they occur.
is
a positive
peak occurring
triggering epoch.
For example, a P300 peak
300
milliseconds after
As figure 3.2 indicates,
of the evoked response are
the
various regions
theorized to correspond to vari-
ous types of mental information processing.
Of particular interest
and latency have
more complex.
by
both memory
tasks.
It thus
is
the P300 peak
been found to vary as
whose magnitude
mental tasks become
Research has shown that the P300 is
tasks and
by
purely cognitive
may eventually serve as a
certain categories of mental loading.
-
26
-
affected
processing
direct index of
-
-T
0
20
'
to
10.4 pV
0
50
410
30
12pV
I
0
11
100
200
300
400
500
1.'.SEC
02PV
0
2
4
12
10
8
6
Figure 3.1:
An -Evoked Response Potential (ERP)
0
0
MSE C
100
"Sensory"
200
300
400
"Processing"
Light Intensity
Pattern Sharpness
Color
Movement
Figure 3.2:
Recognition
Relevance
Expectancy
.
,
5Q0
600
"Motor"
700
Tracking
Components of the Visual Evoked Response
-
27
-
Unfortunately,
workload measurement still
must be overcome.
brain electrical activity for
the use of
has
some practical
The primary problem is that the technicue
as applied to operator workload is
of research.
problems that
still
Three problems that
in an infant state
must yet be
solved are
listed below.
1.
Continuous task and
(e.g.
multiple tasks
piloting tasks) are not pre-
sently amenable to this technique.
mental processes
Different
2.
fere
3.
in
different ways
inter-
with
another and
the effect of
the P300 is
not well known.
one
this on
Signal averaging techniques require
multiple tests of the same event on
the same subject
The great advantage in
that directly
being tapped
using the
reflect cognitive
rather than
of directly
an inference
loading of a subject.
28
loading are
made based
on some
This offers a future pos-
and repeatably
-
that the signals
processing and
variable external to the system.
sibility
ERP is
-
assessing the
mental
3.1.2
Pupil Dilation
The pupil dilation technique seeks
ing
by monitoring
the
expansion
to infer mental load-
and contraction
of
the
pupil, relative to some baseline, as a subject works on some
task.
The technique is based
as information
surge
the nervous
Thus it
processing loads
of 'activation'
Beatty, 1978).
on theory which states that,
are increased,
results
By coincidence,
in
the brain
a forward
stem*
(see
the brain stem also contains
system mechanism that controls
appears that pupil dilation
pupil dilation.
could serve as an index
of information processing load.
An instrument called a Whittaker Pupliiometer can be used
to accurately measure the pupil diameter.
it
has
been found that
increase,
as mental
Using this device
loads induced by
the diameter of the pupil increases
a task
(see Beatty).
Pupil response to task loading generally yields pupil diameter increases of fractions of a millimeter within 10 seconds
of a loading increment.
The pupil dilation
method for assessing mental
have immediate application in
crete cockpit tasks.
load required
analyzing,
This technique
load may
one at a time,
dis-
would allow the work-
for each of the component subtasks in a larger
These neasures could provide
piloting task to be measured.
very useful quantitative workload numbers for use in cockpit
*The brain stem is
the uppermost portion of the spinal cord.
-
29
-
design or flight procedure analysis.
of the pupil dilation
There are two serious deficiencies
method.
This measurement technique requires that the eyes'
dilation due to
ever,
changes
in mental load
be resolved.
How-
the eye also responds to changes in lighting intensity
difficult to detect changes due to
and these changes make it
mental loading.
tially varying
Air transport cockpits are subject to spaso are not
intensities of light and
testbed for the pupil dilaticn method.
a good
Also, new electronic
flight instrument displays (like those of the Boeing 757 and
767) may accentuate this problem.
The pupil dilation
task type environments.
As a direct measure of mental work-
load in an air transport
still
cockpit,
above.
then,
this technique is
may provide useful information
inadequate although it
as was mentioned
in continuous
method is also untried
Compared to the
activity techniques mentioned in
brain electrical
the pupil dilation
3.1.1,
technique may possess an additional advantage.
The techni-
ques of 3.1.1 measure the response to a task or probe exterof interest.
nal or secondary to the task
The pupil dila-
tion method, on the other hand, measures the direct response
of the nervous
system to increases in
As more is learned about the
the primary task of interest.
technique this advantage
processing load from
may prove the pupil
dilation data
more indirect physiological mea-
more valuable than other,
sures of mental load.
-
30
-
variability
Vate
Heart
3. 1.3
Heart rate variability
sinus arrythmia)
is
cardiac or
a technique that has shown some usefulCardiac arryth-
(see Kalsbeek,1973).
ness in basic research
mia is
(often referred to as
calculated by measuring
the duration
between every
from this
computing an
instantaneous
two heart
beats and
heart rate.
estimate of
variance of
The
these estimates
then serves as an expression of cardiac arrythmia.
found lower heart rate variability
Kalsbeek
at higher levels of men-
tal loading.
It
should be mentioned that heart rate variability is
Heart rate was one of
equivalent to an averaged heart rate.
the
heart rate
mental loads.
be an insignificant
has been found to
mental loading.
sive,
for assessing
early methods
not
Most recently Smith
(1979)
However,
index of
did an exten-
full mission simulation where heart rate was recorded
throughout.
(from some
Smith attempted to correlate percentage changes
baseline)
high workload.
rate with other
in heart
These 'other'
indices
number of errors in performing duties,
ciated with tasks, and vigilance.
of workload included
decision times asso-
Smith found no statisti-
cally significant correlation between heart
the three variables.
significant only
acting as
heart
In fact,
with respect
rate and any of
rate was found to be
to whether
the subject
was
during the full mission simu-
'Pilot at Controls'
lation.
-
indices of
31
-
the
sensitivity of
driving simulator
moving base
to a side wind gust
that was subjected
whose magnitude and
The heart rate
location was changed to increment workload.
variability
technique was
significant
relationship
found to
to
technique.
rate variability
the heart
utilized a
Their experiment
evaluated
Wierwille (1979)
assessment techniques Hicks and
their
workload
mental
of
study
comparative
recent
a
In
statistically
have no
operationally
defined
level of workload.
The technique was also found (issues of
significance aside)
to
either subjective
opinion ratings
heart rate variability may
significant workload effects if the
had been
increased,
task perfor-
or primary
In their paper,
mance as an index of mental load.
Wierwille state that
workload than
be less sensitive to
but doubt
have shown
number of test subjects
that this would
relative sensitivity as a workload
Hicks and
change its
index when compared with
primary task performance techni-
the subjective opinion and
ques.
Heart rate variability
index of mental workload for
The fundamental
are influenced by
the piloting
the IFR,
air transport pilot.
problem of the measurement
lies in the fact that heart
method probably
rate and heart rate variability
many factors besides mental
environment.
effort requirements
promise as an
does not show much
These factors
include physical
and physical de.ficiencies,
which may be separable from task related,
data.
- 32 -
workload and
neither of
cardiac arrythmia
Mulder
into its
(1978)
has suggested breaking the heart rate data
frequency spectrum components.
that certain segments of a
Mulder
has found
heart rate spectrum are affected
by the degree of mental loading.
Still,
this has not been
found to solve the primary problem of heart rate techniques.
That is,.
a great many factors affect heart rate variability
other than mental loading.
MEASUFES OF MENTAL
BEHAVIOEAL
3.2
Behavioral measures
the
observation
of
infer mental
seek to
some behavioral
craft/pilot system.
grouped fourteen
WORKLOAD
Williges and
behavioral
output
Wierwille
workload
workload from
of
the
(1979)
measures into
airhave
one of
three major categories:
1.
Primary Task Measures
2.
Spare Mental Capacity Measures
3.
Subjective Opinions
Measures from each of these categories are discussed below.
Piay
3.2.1
It
is
Task Measures
hypothesized that as the mental workload of a crew-
member increases,
measures then,
his performance may degrade.
Primary task
attempt to correlate changes in observed crew
performance with changes
in mental
-
33
-
'wockload.
It should be
noted that the hypothesis underlying the use of primary task
dan (1978),
equivalent,
ter 2).
Simpson and Sheri-
a point of some contention.
measures is
for instance,
that performance is not
contend
nor simply related to mental workload (see Chap-
Nevertheless,
primary task measures have been %sed
in general,
extensively in workload analysis and,
is
found helpful in understanding what
have been
a very complex prob-
lem.
Examples of
measures are
primary task
They
numerous.
dual tracking performance,
include tracking performance,
Generally,
the tabulation of number of task errors.
and
one of
these is related to an operationally defined metric of workload (e.g.
etc.).
or no.
turbulence intensity,
From these relationships
the magnitude of workload
as measured by primary task performance is
In a six degree of
of five workload assessment techniques.
freedom driving simulator they
scales
utililizing the
vals,visual occlusion,
task performance.
used five workload measures.
performance,
lane tracking
then defined.
performed a comparative study
Hicks and Wierwille (1979)
These were:
of turns/second,
of
method
subjective opinion
equal appearing
heart rate variability,
and secondary
Workload was incremented by changing the
magnitude and location of wind gusts on the vehicle.
five techniques,
most sensitive
inter-
Of the
primary task performance appeared to be the
in the
to variations
increments of workload.
-
34
-
operationally defined
?rimary task measures of mental workload have limitations
in
their usefulness for the air transport pilot application.
in
First,
mance
many phases of flight
in the
airport terminal area where
when tracking
he holds this heading,
the localizer course
sonal error and accuracy criteria.
vectored,
pride to
some pilots may consider
vector heading
hold their
another pilot holds his heading +/ledge that the controller will give
as
about the
he might be
approach.
on final
this is that pilots
problem related to
aircraft are
not be concerned
but this pilot may
accuracy with which
For
a pilot will be commanded to hold
generally being vectored,
a heading,
is no simple perfor-
accurately controls.
the pilot
measure which
instance,
there
A
have different per-
For example,
while being
it a matter of personal
2
+/-
degrees,
while
5 degrees with the knowhim a new vector if
the
aircraft drifts off course.
Next,
there is
the problem of cockpit
automation.
In
most phases of an air transport mission at least one control
axis of the aircraft is
example of vectors in
likely set the
tor.
controlled by an autopilot.
the terminal area,
For the
the pilot will most
autopilot heading bug to -the commanded vec-
The auto pilot will then accurately track the heading,
leaving the pilot
free to attend to other
duties.
In the
role of 'flight manager' pilots are trained to delegate more
of the cockpit tasks to
as workloads become high.
the automatic flight control system
With the autopilot in control,
-
35
-
it
may be difficult to specify a
primary tasi measure of pilot
workload.
Finally,
the
most fundamental
problem of
primary task
measures of mental workload lies in the fact that pilots can
and do
exert more
mental effort
level of performance.
pilot flying
to mainatain
An illustrative example of this is
an ILS approach.
Despite the
minor difficulties in the cockpit,
tive is
to
will disregard other tasks,
course with
flight is
or
a
occurrence of
the pilot's major objec-
guide the aircraft precisely down
land if necessary,
a specified
the ILS.
He
place the aircraft in auto-
but he will keep the aircraft on the ILS
strict precision.
A pilot
very likely exerting
in this
increased levels
phase of
of mental
effort but his performance is unvarying.
3.2.2
Spare Mental Canacitgy
Spare mental
capacity techniques
operator is basically a single
CPU computer system.
of processing
That is,
he has some fixed upper limit
capability and
model of
been verified,
it
is
the human
channel instrument or single
tasks on
works absorb some percentage of
hough this
issume that
operator
this fixed capacity.
the human operator
a useful
which the
has not
,in fact,
construct that has served as
the basic foundation of much human factors research.
-
36
Alt-
-
The most common mental workload measures using the theory
component/time summation
of spare
mental capacity are task
methods,
and secondary or side-task measures.
nent/time summation
methods analyze the
ture of an operators task load,
time or total mental capacity
performs
some weighted
tasks or over the total
microscopic struc-
assume a fraction of total
that each task requires,
average over
the
total number
number of seconds.
process results in a workload
total operator capacity.
Task compo-
and
of
This averaging
figure measured in
percent of
Hay, House, and Sulzer (1978) pro-
vide examples of this process.
Task analytic methods have
also been
computer simulations
tied to sophisticated
human operator,
to aid in the
of a
design of cockpit and flight
systems (see for example Wherry,1978;Hay et al,1978).
Task analytic approaches
tations in
such as
these have severe limi-
the rather 'absurd arbitrariness
loreover,
can be associated
task composition
even the accuracy with which times
to tasks,
and the detail
can be defined,
Task component methods,
be
arrive at a percentage work-
applied to the weights used to
load figure.
which must
with which the
is subject
to question.
offer at best an early design
then,
stage measure of mental workload.
Secondary task measures are the second common classification within the
measurement.
spare mental capacity category
These techniques require
-
37
-
of workload
the subject to per-
form a
performance on the side task
ject's
measured as he works
is
The hypothesis
on a major or primary task set of interest.
if
is that,
underlying these techniques
The sub-
called side task).
secondary task (also
the subject's perhe must there-
formance has degraded on the secondary task,
primary task.
fore be working harder on the
-
That is,
the
requiring the use of a cer-
execution of the primary task is
tain percentage of the subject's spare mental capacity.
type of
The specific
secondary task
Examples of side tasks include
varies with the application.
tracking side
tasks,
employed generally
culty varies with side task performance)
and mental arithmetic
ing tasks **,
One
side task
mate 10 to
mary
set
of a song),
results as
an
subjective time estimation (see
the subject to esti-
15 second time intervals while
actively(i.e.
-
tasks ***.
These techniques require
intervals
The
tasks.
of
light cancell-
*,
shown promising
that has
indicator of mental load is
Hart,1978).
tasks (diffi-
adaptive tracking side
perfoming a priare
estimated
have subject tap foot or repeatedly sing bars
or retrospectively (ie.
by comparing the number
an interval with the remem-
and complexity of events within
bered duration of intervals similarly filled).
*See Ogden,1977
**See Ephrath,1975
***See Stephens,
et al, 1980
-
38
-
Time estima-
tion techniques measure the degree
of reliability and accu-
racy of the subject's time estimates while the primary tasks
are performed and compare these with their counterpart baseline time
shape,
estimates.
Hart
(1978)
examines
variations in
central tendency and variability of time estimates as
task demand changes.
Secondary tasks,
though commonly used in
workload mea-
surement have some fundamental problems in their use.
Domi-
nant among these is that secondary tasks are hard to differentiate from primary tasks.
An experinental subject is
very
likely to devote much of his effort to working on the secondary,
rather than
secondary tasks,
this problem.
the primary task.
to some extent,
The
use of adaptive
reduces the magnitude of
As with any neasurement
system,
however,
the
intrusiveness of the measure itself will affect the measured
results.
The intrusiveness of secondary tasks seems to be a
serious limitation of spare mental capacity methodologies.
For the air transport application, secondary tasks suffer
yet another limitation.
task environment.
any given tire in
thus be 'trained'
An IFE pilot works
He may work on
in a multiple
none or on five tasks at
a flight mission.
The secondary task must
into the priority structure
pilot performs other piloting duties.
The introduction of a
secondary task, then, may not indicate the 'true'
of spare mental capacity in
such a t'ime shared,
task envirorment.
-
39
-
with which a
percentage
queuing-type
N.
3.2.3
Subjective Opinions
Subjective opinions
of mental workload in
are perhaps the most
use.
Sheridan
that mental workload be defined
jective
Indeed
experience of
as
his or
most experimenters
'a
her
do
common measure
(1980)
would suggest
person's private sub-
own cognitive
calibrite their
effort.'
'objective'
measures against a subjective scale or interview.
other measures discussed in
ions are generally
this chapter,.
Like the
subjective opin-
used in conjunction with
other workload
measures and have been historically useful in this regard.
Subjective opinion techniques utilize two basic formats:
1.
Interview/Questionnaire
2.
Rating Scales
Interviews and
questionnaires are generally
tured, information gathering devices.
workload metric
scales,
is thought to
on the other hand,
implemented according
psychometric theory
to the
for assessing aircrew
according to
be rather
linited.
Rating
body of
literature known
as
(see Nunnally,1967).
rating scales offer one of
of the
Their usefulness as a
say be rigorously structured and
Gartner and Murphy (1976)
none
loosely struc-
state that subjective opinion
the most promising methodologies
mental workload,
scales which
they
surveyed were
the proper techniques
-
40 -
but note
of attitude
also that
constructed
scale con-
Edwards.,1957).
struction (see
Such
to rating
or are subject
unanchored,
ambiguously worded,
are
scales generally
intransitivities.
The Cooper-Harper
is
craft handling qualities,
scale in
used by test pilots to rate air-
scale,
perhaps
the best known rating
of aircraft
far the test and development
use
scale that
a workload rating
Figure 3.4 is
figure 3.3).
(see
appears to have been modelled after the Cooper-Harper scale.
more
like a
qualities/performance
handling
physical components of
the
scale than
a
the scale seems to emphasize,
Also,
workload rating scale.
primarily,
too similar and the scale reads
the model is
Unfortunately,
workload,
ignoring
mental contributions.
a multi attribute scale for
Figures 3.5a and 3.5b depict
measuring the workload of an IFR pilot that was developed at
MIT.
This scale
(1S78)
theory developed by
perceived busyness,interruptions,
occurrences of
tasks (see Chapter 2.).
reason why
stress,
3.5b,
Figure
a specific rating
and simultaneous
and planning
in turn,
(from 3.5a)
is
weighs the magnitude of fraction of time busy,
information processing,
transport
3.5 uses the concepts of
monitoring,
operating,
Simpson and
the air
on mental - workload in
The scale of figure
environment.
the Cooper-Harper
also modelled after
is based on the
scale but
Sheridan
is
type
seeks the
chosen.
It
intensity of
and the intensity of feeling or emo-
-
41
-
Satisfactoy
Meets all requir-
Acceptable
ements and expectations, good
enough without
improvement.
may have deficiencies which
warrant improvement, but adequate for mission.
Capable of being
controlled or
managed in context of mission,
with available
pilot attention.
Clearly adequate
for mission.
Excellent, highly desirable
All
Good, pleasant, well behaved
A2
Cai.
Some0
m.
I
l'
y
unp
l
easn
c_
it
t
h
t
"aa
er
s
Good enough for mission without improvement.
Pilot compensation, if required
4
to achieve acceptable performance Unsatisfactory
is feasible.
Reluctantly acceptable. Deficiencies which
warrant improvement. Performance adequate
for mission with
feasible pilot
compensation.
Some minor but annoying deficiencies. Improvement is requested.
Effect on performance is easily compensated for by pilot.
Moderately objectionable deficiencies.
IA3
A4
Improvement is needed.
Reasonable performance requires considerable pilot compensation.
AS
Very objectionable deficiencies. Major improvements are needed.
Requires best available pilot compensation to achieve acceptable
performance.
'A6
I
I
Major deficiencies which require mandatory improvement for
acceptance. Controllable. Performance inadequate for mission,
or pilot compensation required for minimum acceptable performance
in mission is too hiqh.
Unacceptable
Deficiencies
which require
mandatory improvement. Inadequate
performance for
mission even with
maximum feasible
pilot compensation.
Controllable with difficulty.
U7
Requires substantial pilot skill
and attention to retain control and continue mission.
Marginally controllable in mission. Requires maximum available
pilot skill and attention to retain control.
*U9
4
1
Uncontrollable
Control will be lost during some portion of mission.
Figure 3.3:
Uncontrollable in mission.
Cooper-Harper Scale
Ui(
OEM'A.NOS CN. THE PItO T IN SELECTED
TASK OR REQUIRED OPERATION
PItOT
RATING
PtTEFFORT NOT A FACTOR
FOESIRED PERFORMAACEI
WORKLOAD
REDUCTION
NIOT
tAINIMAL PILOT EFFORT REQUIRED
N-ECESSARY
FOR DESIRED PERFORMANCE
YES
DESIREA PERFORMANCE REQUIRES
MODERATE PILOT EFFORT
IS WORKLOAD
LEVEL
SATISFACTORY?
WORKLOAO
AREDUCTION
WARRANTED
AOEQUATE PERFORMANCE REQUIRES
CONSIDERABLE PItOT EFFORT
ADEQUATE PE-RFOR.M'ANCE REQUIRES
EXTENSIVE PILOT EFFORT
A
IS ADEQUATE
PERFORMANCE
ATTAINABLE WITH
A TOLERABLE
WORKLOAD?
NOT
PERFORMANCE
ADEQUATE
ATTAINABLE WITh MAXIMUM TOLERABLE
PILOT EFFORT. CONTROLLABILITY
NOT IN QUESTION
WORKLOAD
REDUCTION
REQUIRED
FOR
ADEQUATE
PERFORMANCE
CONSIDERABLE PILOT EFFORT IS
REQUIRED FOR CONTROL
INTENSE PILOT EFFORT IS REQUIRED
TO RETAIN CONTROL
YES
IS AIRCRAFT
CONTROLLABLE?
5
V
.
REA
L
CORKLOADN
CNT RY
DANDATORY
,
CONTROL WILL BE LOST DURING SOME
PORTION OF REQUIRED OPERATION
PILOT DECISIONS
Figure 3.4:
A Workload Rating Scale
6
Chapter 6 of this thesis describes a flight simulator
tion.
figure 3.5.
evaluation of the rating scale depicted in
methodologies have limitations.
ing,
skill,
to vary with the trainSecond,
and emotional state of that subject.
there is
the problem that
perceive
his own
the subject
That
workload.
is, he
thdt
so
'hard' the test
and to consistency within and
related to scale sensitivity,
across subjects
may become
there are questions
Finally,
after the fact.
able to
may not be
duties that he forgets how
immersed in his
interval was,
For instance, the workload
likely
rating a subject provides is
subjective opinion
measures,
Like other mental workload
limitations on the
may impose
use of
subjective opinion rating scales.
construction techniques,
Proper scale
training
of sub-
jects in the use of the rating scale and a structured experimental scenario
problems.
magnitude of
should lessen the
Hicks and Wierwille (1979),
found that a
second most sensitive indi-
subjective rating scale was the
cator (next to a primary task measure)
mined.
in fact,
the above
among five they exa-
Hicks and Wierwille utilized the
The scale used by
method of equal appearing intervals.
Aside from
their sensitivity,
subjective
opinions have
another advantage in that they are almost completely non-intrusive to the test.
As such, these- techniques are far simtask or physiological type
pier to implement than secondary
-
4 4
-
methodologies.
to a much
Subjective methods,
then,
lesser extent with the final
should interfere
or measured outcome
of a workload experiment than other assessment techniques.
-
45
-
IFR TRANSPORT PILOT WORKLOAD SCALE
Second Version, May 1979, MIT FTL/MML
Start
Is
Workload
Satisfactor
Yes - Satisfactory
no
See
Table
Low Levels of Workload -- Such
That All Tasks are Accomplished
Promptly
0
S
I.
Is
Workload
Acceptable
Yes - Acceptable
?no
Is
Workload
Possible
Moderate Levels of Workload
Such That the Probability of
Error or Omission is Low, But
Improvements are Desirable
High Levels of Workload --
Yes, But
Unacceptable
Such That the Probability of
Errors or Omissions is
Unacceptable
no
Impossible
Figure 3.5a:
It is Not Possible to Perform
All Tasks Properly
MIT Scale, Part I
See
Table
A
4
THREE ATTRIBUTE RATING OF PILOT MENTAL WORKLOQ (continued)
FRACTION OF TIME BUSY
seldom have anything to do
often
occasionally
have free moments of time:
-
very rarely
fully occupied every single instant
2.
INTENSITY OF THINKING/INFORMATION-PROCESSING
activity is completely automatic; no conscious thinking or planning required
cerebral effort and planning required due to problem
low level, occasional
complexity.. uncertainty., unpredictability.,
unfamiliarity, etc., is:
moderate
high level
supreme mental effort and concentration are absolutely necessary
3.
INTENSITY OF FEELING
experience is relaxing, nothing to be concerned about
emotional stress, anxiety, worry, frustration,
confusion, etc., are:.
mild, occasional
moderate
Ihigh level
severe and intense psychological stress
Figure 3.5b:
MIT Scale, Part i
3.3
S LMMARY
It
can be
measure of mental
the air
as yet,
truthfully stated that there is,
workload that is
for use in
well adapted
measures cur-
of the
All
transport environment.
no
infant state of
research and
far from well-understood, production type use.
Physiologi-
rently under
cal measures
study are in an
directly
related
to the
These techniques are,
techniques
tapping signals
indices of mental load,
promise as 'hard'
show great
electrical activity
such as brain
brain's
information
the
however,
processing.
furthest of any of the
surveyed from production type use.
Behavioral measures
may be that,
of mental load
are numerous
of task demands,
for a small set
and it
or for some
currently available behavioral metho-
range of workload,
a
dology is useable.
These techniques are limited by the fact
that many of the behavioral methods are based on human operator theories
shown to be
that have been
inaccurate.
An
of primary task measures of men-
example of this is the use
tal load despite the fact that
pilots can and do,
on occa-
sion, exert increasing levels of effort to maintain a specified constant level of performance.
Two behavioral
show
promise for
workload assessment techniques
use
in
These are subjective time
rating scales.
the air
transport
appear to
application.
estimation and subjective opinion
Their use also awaits the results of further
-
48
-
research,
but results
to date indicate that
these two are
viable techniques.
It
cannot be stated,
at present,
one universal rreasure of mental load,
application.
iety
there will ever be
even for the piloting
As in the past, it is more likely that a var-
of assessment
applications and
stated
if
that the
will
techniques
workload ranges.
be used
It can,
available uethodologies
for
various
however,
will be
improved
over those presently available to the aviation community,
theory and experimental research in
more mature stage.
-
49
-
be
as
this area progress to a
Chapter IV
PILOT WORKLCAD AND THE INELU-ENCE OF AIR TRAFFIC CONTROL
The
literature deals
at length
with
the analytic
and
theroretic description of piloting tasks and their affect on
pilot workload.
The cockpit,
however,
is
not an isolated
unit consisting of pilots and flight controls.
is
a component
pilots,
of
a larger
the aircraft
system
The cockpit
that encompasses
controls and dynamics,
way/air traffic control system.
and
the
the air-
Workloads are imposed not
simply due to physically operating the aircraft but also due
to
the air
itself.
traffic control
For example,
and
National Airspace
an instrument approach is a series of
turns, descents, and speed control maneuvers,
of which require
some physical and mental
approved approach
the execution
effort.-
procedure further constrains
of maneuvers to be executed
tical limits
System
and with a
craft's approach speed ,
An FAA
this series
within certain lateral and ver-
certain accuracy.
specific
Given
an air-
tasks in the cockpit are
triggered by the constraints of the approach procedure.
'approved' approach
what is
procedure is
an integral
The
component of
referred to in this thesis as the 'ATC system.'
The cockpit procedures and the air traffic control system
are interrelated and the interface
-
50
-
between these two compo-
nents directly affects the pilots workload.
above both
the system structure
and the
could be changed to effect a change in
tion of pilotinq tasks,
standing workloads in the
explores the
mechanisms
under-
air transport environment.
This
in
some detail.
the cockpit
interact are discussed in relation
procedures
sis.
Next,
the timing and execu-
an important issue in
issue
through which
of Chapter 2.
cockpit procedure
and thus workload.
The ATC/cockpit interface is
chapter
In the example
and
First
the ATC
the
system
to the workload concepts
several tools for quantifying cockpit
are described that are
useful in
workload analy-
The tools are particularly useful in that they specif-
ically address
chapter a
the ATC/cockpit relationship.
routine arrival
into Boston's
analyzed using the concepts introduced in
In
the next
Logan airport
is
this chapter.
LOADIVG 8ECHANISMS
4.1
There
are
several
mechanisms
traffic control system can be
load.
through
which
the
seen to influence pilot work-
These mechanisms can,
for convenience,
be grouped
into three categories:
1.
Airspace mechanisms
2.
Regulatory mechanisms
3.
ATC voice ccmmunication mechanisms
-
air
51
-
The airspace
structure,
Federal Aviation
presence of
the
Regulations, and the character of air traffic control communications are assumed to be integral parts of what is referred to in this thesis as the ATC system.
THE AIRSPACE LOADING MECHANISM
4.2
The airspace
that
is invoked
pilot loading
loading mechanism relates to
by the
the airspace
struture of
system.
Examples of cockpit tasks that are cued by the airspace system are numerous.
1.
Task:
Change
radio
Required as aircraft
frequency.
passes an air
or bend in
traffic sector boundary
an airway.
2. Task: Climb to an enroute altitude.
Aircraft generally must maintain an
altitude
less than
cruising
altitude until
craft
imposed
is
clear
by the
planned
their
of
the
air-
restrictions
airspace
terminal
structure.
3.
Task:
Instrument approach
proce-
dure.
There
exist
defined
well
lateral and vertical limits of protected
airspace within
which
the
approach maneuvers must be executed
-
52
-
(see
Terminal
Enroute
These
Standards).
Procedure
limits
constraints which tend
all
to restrict
tasks
landing/approach
a small
executed within
impose
be
to
time win-
dow.
also define
which
procedures
approach
Instrument
navigation fixes
approach related
Figure
executed.
4.1
tasks
are
provides
an
physical cock-
example of specific
pit tasks that
over
are associated with
navigation fixes.
Summarizing,
by
constraining or
loading mechanism operates
the airspace
invoking tasks.
tasks are
affected by
workload,
ty definition
the system
Because the
pilot's
be the
pilot's
so must
(see Chapter
2).
of the U.S.
tasas evoked by common elements
Table 4.1 lists
National Airs-
pace system.
4.3
PEGULATOEY LOADING MECHANISMS
Certain
aspects
with the U.S.
into law.
and operating
restrictions
associated
system have been regulated
National Airspace
There also exists a large body of nonregulatory,
but recommended set
of operating procedures that
-
53
-
have been
defined to
standardize aircraft movement within
for air traffic control purposes.
the system
This body of rules tends
to restrict and predefine piloting tasks, and thus loads the
pilot above
the conditions
encountered in
an uncontrolled
environment.
Regulatory loading mechnanisms are a
mechanism discussed in
pace loading
however,
subset of the airsIt
section 4.1.
convenient to separate out this subset for organi-
zational purposes.
Table 4.2 lists the documents which spe-
cify rules and operating procedures for aircraft.
be
noted
that
although
some
nonregulatory information,
tion stresses
Table
is,
common
the
documents
should
contain
the Federal Aviation Administra-
adherence to
4.3 lists
of
It
these procedures,
tasks
invoked by
the
nontheless.
regulatory
constraints, as defined by these documents.
A simple example will serve
to illustrate the regulatory
loading mechanism and its affect on cockpit activity.
ure 4.2 is
an annotated schematic of
into Presque Isle, Maine.
Fig-
an instrument arrival
Presque Isle, though not typical
of most domestic transport operations, is presently a terminal into
which Delta
Airlines schedules
arrivals.
-
54
-
three Boeing
727
I.
VOR C
CONTROL SECTOR A
VOR Changeover Point
o Switch Nav to VOR C
CONTROL SECTOR B
030
V 14
VOR B
VOR A
o Change Nav Receiver
o Turn Left 60 degrees
Compulsory Reporting Potint
o If non-radar environment,
report position and give
estimate to next fix
Figure 4.1:
ARTCC Sector
Boundary
Flight Along Federal Airways
o Change Communications Frequency
o Transmit Messages
Table 4.1:
Airspace Task Triggers
Airspace Component
Associated Tasks
VOR
o
Turn
o Navigation Frequency
Change
o VOR Course Selector
Change
Airway
o Turn
Intersection
(Bend in Airway) -
o Reset Course Selector
'-N-
VOR Changeover Point
o Change Navigation
Frequency
Control Sector Boundary
o Change Communication
Frequency
o
Ident Transponder
o Transmit Message
Terminal Area Boundary
o Climb/Descend
o Accelerate/Decelerate
to Designated Speed for
Air Traffic Control
-
56
-
I .....
Table 4.2:
Documents Containing Flight Operating Rules
Document
Federal Aviation Regulations
Contains
o
Part 91 - contains
general
rules pertaining to all
aircraft
o
o
Part 135 - contains additional
operating rules for scheduled
air taxi operations
Part 121 - contains additional
operating rules for scheduled
air transport category operations
Advisory Circulars
o
Contains nonregulatory advisories
and practices which pertain to
specific parts of the Federal
Aviation Regulations
Airman's Information
Manual, Part I
o
Contains nonregulatory information regarding recommended
operating practices and other
general information concerning
operations under Instrument
Flight Rules.
Table 4.3:
Common Tasks Invoked by the Regulatory Mechanism
Trigger
Tasks Required by Law
Descend Below 10000ft
o Slow to less than 250 knots
Cross 18000ft
o Change altimeter setting
Hold at a fix
o
Slow below maximum authorized
holding speed (as specified in
Airman's Information Manual,
Part 1.
o Fly holding maneuver as per
AIM, Part 1.
Cleared for the approach
o Fly approach procedure as
prescribed in Federal Aviation
Regulations, Part 97.
-
58
-
134
179
314
359
A
79
E
G
Procedure Turn Must
Within lOnm of VOR
I
C
,
be Executed
2500ft
4*
I
1260ft
1~1
I
2.5nm
Figure 4.2:
140ft
G
2.3nm
The Presque Isle Approach Procedure
From figure 4.2 and Table 4.4 it
is
clear that Ethe proce-
dural requirements of the approach constrain certain cockpit
tasks
executed
to be
entirely clear,
at predefined
however,
to
what extent these constraints
approach.
affect total workload during the
full impact of
7o measure the
regulatory and airspace constraints
on crew
a chrono-
workload it is necessary to enumerate, in detail,
logical record of cockpit activity during the approach.
record
of cockpit
activity
not
It is
points.
is
referred
to
as a
This
cockpit
activity timeline.
Figure 4.3 is a cockpit activity timeline*
727's ajproach to runway 19.
In constructing the timeline,
a priori, some assumptions were necessary.
was assumed to begin its
approach
assumed to be at
The 727 aircraft
over the Presque Isle TOR
at 2500 feet, on a heading of 359 degrees.
craft was also
for a Boeing
The subject air-
the speed and in
the air-
craft configuration specified by the Delta Operations manual
for the initial
approach segment.
the timeline are taken from
Pilot's Operating Manual,
the figure,
manual.
The procedures shown in
the Delta Airlines,
Book I.
were estimated using
Boeing 727
Cockpit event times,
on
speeds called for by this
Wind velocity was taken to be zero.
*For readers unfamiliar with timeline analysis,
sion is included in the appendix.
- 60 -
a discus-
The task
that, in
timeline for
the Presque
Isle approach
addition to the tasks that are defined
shows
in Table 4.4
by the procedure, other tasks associated with the safe operation of the aircraft must be squeezed in between the procedural constraints.
For example,
the timeline indicates that the
on Panel 3 of figure 4.3,
checklist is
performed by the first and second officers.
must generally be completed in
the VOR is reached.
If a
landing gear problem) ,
being read and
This checklist
the 100 second period before
difficulty should occur
would have little
the crew
(e.g.
a
time for
decision making in this high task load situation.
Panel 4 of the timeline indicates another problem area of
According tc figure 4.2,
the procedure.
VOR the pilot must step down,
2,5 miles later,
first
to 1140 feet.
upon reaching the
to 1260 feet and then,
The timeline shows that the
second step down begins only 20 seconds after the first
has been corpleted.
Moreover,
140 knots throughout
the phase,
the pilot is
if
he will
applying speed control techniques in
nuous
Again,
tracking
tasks
the nature
of the procedure
tively higher task demand
situation.
that at this higher loading,
the cockpit,
there is a
if
to maintain
be intermittently
addition to the conti-
quick
and the
step
has
step-down
descents.
invoked this relaTheory would suggest
difficulties should arise in
greater probability that the pilot
will make an error.
-
61
-
Figure
4.4 is
useful
for
identifying the
which the Presque Isle approach
of the arriving 727.
constraints
imposes an cockpit activity
The boxes of figure 4,.4 represent dis-
crete events and are positioned along event paths with which
they can be associated.
left to right
Time,
on the figure,
increases from
to the left of
so that an event
another
(on
the same path) is constrained to precede that task.
A solid
line connecting two
events indicates
task is being performed between these
seconds between
parentheses.
such events is
A dotted
That is,
constraint.
ween these two events
that some
two and the number of
indicated by the
line indicates
a soft
number in
precedence
there is no task being performed betbut the event on the left must pre-
,
cede the event on the right end
of the dotted line.
It is
implied that the time between two events which are connected
by a soft precedence constraint is
zero seconds.
(See the appendix
greater than or equal to
for a detailed explanation
of task precedence maps in wcrkload analysis)
The task precedence map of
capability in
certain events
example
rence
displays the
precedence with
trigger certain'other cockpit
(see Panel 1 of figure 4.4)
to precede
raint.
that it
figure 4.4 possesses a unique
event D1
by the
events.
which
For
event Al is constrained
dotted soft-precedence
const-
Crossing the VOR outbound, event Al, cues the occurof event
Dl,
the
beginning- of outbound
-
62
-
tracking.
-
Event Al also cues event D2,
in turn, cues event C1,
event B1,
is
D8,
The implication is
other
shows that almost all events
and D9.
Event
except event A2,
D8 then cues
that one position event,
events.
Further
are triggered
A4,
examination
(or constrained)
by events along path A, the position event path.
now to figure 4.2 it
Event D2,
Panel 2 of the figure indicates that
tasks C3,
constraining five
.MWANAM"
the extension of 5 degrees of flap.
As another example,
event A4 cues
a slowing maneuver.
----
Referring
can be. seen that all events on Path A,
are predefined
by the instrument approach
procedure itself.
-
63
-
I
Table 4.4:
Presque Isle Approach Tasks
Location
Major Control Tasks Triggered
A (at VOR)
o Becin outbound timing (1 minute)
B (1 min. from A)
o Turn to 314 degrees
C (45 sec. from B)
o Turn to 134 degrees
D (intercepting
o Descend to 2000ft
"finaI")
o Turn to 179
E (VOR) -
o Descend to 1260ft
F.(2.5 DME)
o Descend to ll40ft
F (4.8 DME)
o Make GO/NO GO Descision
(
7T- 7r77to 314
7
HEAD ING
-H359
TASKS
-----------
desl
{
_ 14 d e- g)
-'
m--------------------m-----
ALTITUDE
H2500f
-H25 0 0
TASKS
SPEED
Se-ec-t 15F--E
L-t---
)Se
5 FIa
Iectp
2) SlI ow
TASKS
to
4Hi
4 Slow
16)i4 6010
to
-
I
k t
150
AND
NAVIG.
PLANNING
TASKS
'.----.---.--------.---.--------------------COMMUNIC.
TASKS
5|3) P
2) P2-:Fla
20
..
,.. --
40-
: Flap_
4
1
2 P2 Fla
60
l
100
80
__I
Figure 4.3:
Cockpit Activity Timeline, Presque Isle
4
P1
t
*
9,
0
(
TLl
HEADING
TASKS
ALTITUDE
TASKS
SPEED
TASKS
iTLl
NAVIG.
AND
PLANNING
5 No Smoke-On
2) Set Alt.3 Se t A/S Bug
4)Set EPR Bug
Bu s
TASKS
COMMUNIC.
TASKS
\
6)P3: e-ugs
t Alt.|13)P2:Gear
-P-3: A/S and ~
EPR Bugs'
10) P3:No
12)P1 :Gear Down| Flap 25 l
SmkJL_
P2:Set
9 7 P2:Set
220
I lP2
240'
Down
FI ap 25
)
r
B
Lodg
Cpte
:On
260
280
300
1
0 0-
TASKS
onw
NAVIG. AND
6)Missed Apprch
Decision
PLANNING
-
TASKS
COMMUN IC.
115JP2:1260ft]
tL6)iiijinrd
TASKS
-* a
320
.........
310
360
4(
380
PA
- -
..
TL 3,4..
-
-
-
-
NG
4) rack
179
RadialI
'UDE
4)Maintain
-
2-
ll40ft
T
2
TL
2-....
..
.....-
....-
- -. --
-
-
-
--
--
-
-
-
-
TL 2-6
3.AND
qING
5
m
i
______________
0P5
m-. ........-
m-
mTL
.
-..-.
-.
-..-
-.
-
--
-
-
----
--
JN IC.
17)
P2 MAP
420
44 a0
460
500
480
P5
Pos it ion
Eve ts
(85)
A
Altitude
Events
&
I
0
I
*
Communications
Events (Interc
C2\
--
115
deg
-------
Il a
60,430)\
Pl ann ing
Eve ts
E
(0,333)
Figure 4.4:
--
-
..
...
.
-
.
(4
Task Precedence Map,
.
..
Presque Isle
--
.
.
.
-
.E
(
I
(
(0)
(205,205)
Bl
-
Leave
2500
B2
2000
ft
/
/
- - -
-
Loc
1
~izer~
205,
f t
333)Z
\(205,
(D ----
Arrive
--
-
30)
\
/
he ckKD
I
list
(205,430)
/
Dl 1
Begin
Level
off
-
-- ff
(205,358)
G) - - - -- -
.
-
3)E
-
, 370)
(300, 300)
5
1B
B4
Arri ve feave
B3
12600 ftl I
200
00,937
370,418)
(345,418)
B6
Arrive
1140 ft
.382,430t\
(C)-\ ~(382,430)
O
D14~
D1
DBegin
Descent
_
(300,373
egin
Begin
Leveloff
_j
30,,
3709418)
(30,15)
1~EJ------------------------
D
Begin
. _
Descent
Leveloff
D16
D15
]~Begin
-
-
-
-
----
----
---
-I'issedE
(382 ,430
j)
4.3.1
Discussion of the Regulatory and Airspace Mechanism
It has been
argued in this chapter that
control system
has a
strong influence
rate of the cockpit crew.
the air traffic
on the
task demand
The discussion to this point has
shown how the structure of the airspace system tends to constrain and invoke
Reiated to
cockpit tasks.
exists a body of rules,
this,
there
which instrument pilots must follow,
that have been instituted to control the movement and behavior of aircraft in the system.
ATC system has an affect on
port environment.
be
that it
is
There is no doubt that the
crew workload in
As cockpits become
the system
the air trans-
more advanced it may
structure,
not the
operating
requirements of the aircraft , which induces the predominant
pilot loading.
Some further discussion, though,
is neces-
sary on the nature of this ATC/cockpit interaction.
An important factor in the ATC/cockpit interface that has
been omitted in the previous discussion is flight operations
procedure.
Flight
procedures also influence
the workload
level of the crew,
though they are designed to minimize such
loading effects.
An example of
poor procedure design will
clarify this point.
Suppose that' an aircraft's
that full
glideslope.
flaps should
operations manual
be extended
specifies
upon intercepting
the
Suppose further that the subject aircraft type
tends to pitch nose-up as full flaps are extended
-
7' A
W
(as is
the
This procedure induces a relatively
case on many aircraft).
time of glideslope intercep-
high workload situation at the
The high workload would result from the pilot having
tion.
pitching moment while
for the upward
to compensate
transition down-
effecting the aircraft's smoota
same time
craft operations
should be noted that most air-
It
wards onto the glidepath.
flap extension
manuals call for
glideslope interception
at the
prior to
(see Delta Airlines Boeing 727 oper-
ating Manual and American Airlines B707 Operating Manuals).
through the
cockpit task
demand
triggering method
as the
affect
also
procedure can
Flight
same constraint/task
airspace and regulatory mechanisms.
ure 4.4,,
event D1 is
in
the example of fig-
seen to trigger event D2 and this then
It
triggers event C1.
In
is
the flight procedure as specified
the operating manual that imposes these constraints.
Most often
the design of airspace
or the design
coordination with
performed in
traffic control procedure is
the design of safe flight procedure.
of air
Nevertheless some fed-
eral procedures have been implemented which are questionable
in
Noteable in
their workload implications.
the steep,
decelerating noise
this regard are
abatement arrival procedures
which have been implemented at some terminal areas.
It
is
difficult to conceive of the National Airspace sys-
tem changing
For
or of regulated procedures
this reason
it
may seem
-
74
changing markedly.
unimportant
-
to analyze
the
of
influence
Despite
workload.
influence.
this,
useful to
is
it
system
control
traffic
air
the
on
crew
recognize this
Understanding the complexitites involved in the
workload of an air transport crew can only help in designing
Also as systems such as
systems that are workload tolerant.
RNAV
and ATC
point navigation)
(point to
DABS radar are implemented,
systems based on
system will indeed change.
the
will be important to have analyzed the airs-
At that time it
pace and regulatory loading mechanisms.
4.4
ATC VCICE COMMUNICATICN
4.4.1
MECHANISMS
Explicit Loading
There are
essentially two
which air traffic control
different mechanisms
through
voice communications affect pilot
workload.
First, there is
which is
Every ATC message
an explicit mechanist.invokes,
directed to an aircraft
as a minimum the
tasks:
In addition
1.
Grasp microphone
2.
Respond to message
more complex tasks,
clearances
directed to
most messages
an aircraft
invoke
These messages come in the form of ATC
and advisories.
Table 4.5
lists some
piloting tasks that are triggered by ATC messages.
-
75
-
common
discussed
the airspace mechanism
gered by
For instance,
section 4.2.
in
must reduce their speed
inbound aircraft
all
often trig-
invoke piloting tasks are
ATC messages that
to 313 knots or less when crossing a particular sector boundary,
in
and is
unaware of this
is
however,
pilot,
the terminal area.
order to reduce delays in
in
Table 4.5)
tion task,
the sector boundary
is
that defines where
airspace structure)
(see no.
2,
to execute the speed reduc-
causes the pilot
but it
also unaware of
A message
the location of the sector boundary.
The
(a component of the
this task
must be
initiated.
The ideas of interruptions, constraints, and triggers are
important concepts
Figure 4.5 is
visualizing
another.
It
in
the
constraints
is
also helpful
map is
especially
that
imposes
event
on
identifying when one event
in
In
one
useful for
is
which
a task precedence map*
triggers another to occur.
dence
Chapter 3.4).
workload analysis (see
this context the task precefor identifying
useful
how
air
traffic control messages cue piloting tasks.
box C2 to box
The dotted precedence constaint connecting
E2 indicates that event E2 is
C2.
The Begin
cued by the ATC message of box
Slowing event of box
communications event,
Cl.
However,
also cued
event C1 is
Summarizing,
Al.
by a sector boundary event,
Dl is
76
-
in turn cued
as the subject
*See section 4.3 for a brief explanation of
maps, or the appendix for more detail.
-
by a
precedence
9
boundary the controller initiates
aircraft crosses a sector
in turn evokes a
command which
a speed control
task
in
the
Figure 4.5 thus highlights the difference between
cockpit.
a task invoked by the airspace mechanism and task invoked by
the communications mechanism.
4.4.2
Implicit Loading
This is
load the pilot.
sages tend to
through which
a second mechanism
There exists
ATC mesto as the
referred
implicit ATC loading mechanism.
the pilot
whether they are directed
a transmission occurs.
tion from other tasks when
(see Morgenstern)
area,
These
experiments found
could
recite the
aircraft,
frequently
transport pilots
identification and
altitude of
multiple
both preceding and trailing the subject aircraft.
An example will
mechanism.
transmissions.
control voice
that
Experi-
the terminal
traffic situation in
air traffic
based on
This
that pilots construct a
have shown
of the air
mental image
not.
to him or
pilots divert some atten-
occurs because trained instrument
ments
frequency load
the air traffic control
Transmissions on
illustrate another form of
the implicit
arriving at
the terminal
Suppose
area expecting
an ILS
a flight is-
to runway 4
the message,
with
-
presently about
Suppose now that the control-
fifteen minutes froniw landing.
ler contacts this aircraft
and is
77
-
'Amend clear-
Table 4.5:
Explicit Loading Due to ATC Messages
Message
1) "Descend
to 18000"
Type
Tasks Invoked
Clearance
o Descent Maneuver
o Descent Checklist
2) "Reduce Speed to
Clearance
o Speed Control
310 knots."
3) "Turn left 080,
intercept the ILS
Runway 4R."
Maneuver
Turning Maneuver
Navigation Maneuver
Clearance
and approach checklists
4) "Reports of moderate
o Planning Request
Advisory
turbulence at FL240
for New Altitude
by a 707."
or for Speed Reduction
to Penetration Speed
5) "Descend pilot's
discretion to 8000ft
Clearance
o Plan Descent Rate
and Top-of-Descent
Point to Allow idle
Thrust Arrival
6) "Expect 10 minute
Advisory
o Locate MILlS intersection.
delay at MILIS
intersection."
o Slow enroute or prepare to hold.
-
78
-
s it ion
A2
-
en ts
CVOR
-...-
(1 80)
-ro-
Cnrosus
ISector
t itude
en ts
Leave
18000 ft
V
R
0...
(0)Leave
-....
1000f t
mmun i cat i on s\
(160)\
C1
ents
to 310\
knots\
anual
Control
ents
*DI
D2
i
BeginBe
Slowing
Slowing
1250 kts
IManeuvr
ning
ts
El
Begin-
-
-
Figure
-
-
4.5:
-
-
Communications
-
-
-
-
vs.
-
Airspace
-
-
-
Mechanisms
-
-
-
-
-
-
ance.
This clearance
Expect VOR-DME approach to runway 27.'
causes a large planning task to
short period of time.
be executed in
a relatively
Such a clearance can also implicitly
arouse anxiety
introduce confusion and
in the
cockpit
chapter 3 for the effects of anxiety on mental load).
this type
of ATC
induce large
message could
upon the crew even though
(see
Thus,
mental loads
the associated physical tasks are
rather minor.
In fact,
All ATC transmissions mentally load the pilot.
it
is
hypothesized that at some
workload level a pilot can-
not control the aircraft within strict tolerances and interThe effect of radio
pret ATC messages at the the same time.
messages on
pilot workload is
representation
of figure
workload, on an ordinal,
the schematic
summarized by
The
4.6.
figure suggests
subjective scale is lowest
that
in the
and higher in the terminal
long descent phase of an arrival
area vectoring and final approach portions.
These are work-
load ratings measured with the vectoring and altitude clearances as the only ATC communications.
Overlayed onto
plot of ATC
rating of
the workload
(Channel Util.
channel utilization
Time/Total Air
Time,
figure 4.6
over 20
averaged
is
a
= Busy Air
second intervals).
The channel utilization curve was sampled during a peak hour
on an IFR day at Boston, Massachusetts.
The data represent
samples from one center sector, one transition approach sec-
-
80
-
ATC Channel Utilization
33%
30%
Workload
Level
"Final"
Workload
Level
Level
TIME
Figure 4.6:
ATC Communications and Pilot Workload
C
0
co
1)
0
TERMINAL
aVECTORING
DESCENT
FINAL
'4-
-4
0
Z0
100
500
1000
1500
Time into Mission(sec.)
a) Smoothed Communications Message Rate
4)
0~
'U
L.
4)
'U
0~
2
I
4)
U,
0
00
I-
DESCENT
TERMINAL
.... 4|VECTORING
F1NAL M
In
100
' 1000
500
Time into Mission(sec.)
b) Communications Channel Utilization
Figure 4.7:
High Rate Profile for a 'Busy'
-
82
-
LAX Arrival
tor,
and one local tower control position.
sesses a
shape common
to communications
The curve posactivity at
many
terminals in the U.S.
Figure 4.6 indicates that
terminal area vectoring
arrival is
the channel utilization during
and final approach segments
double that for the long descent segment.
over, in this sample,
is slightly higher
than that during the
vectoring portion.
that although,
the nominal workload due
final approach
More-
channel utilization on final approach
What the-figure suggests is
pilots,
of the
may be low,
for experienced
to flying the aircraft on
the mental loading
induced by
high ATC channel utilization may add a significant increment
to the total pilot workload level.
during the long
cantly affect the
High channel utilization
descent phase of an
arrival could signifi-
measured workload on what
low workload segment.
Figure 4.7 is
munications profile during a 'busy'
is generally a
an example of the comvisual arrival into Los
Angeles.
Air traffic control voice communications
crew.
load the cockpit
This loading is generally applied in three ways:
1.
or
Clearances
advisories
execute some physical
the pilot to
or mental piloting task.
-
require
83
-
C
0*
0
1
2
3
4
5
6
7
Transmissions/20 second interval
a) ILS-DME APPROACHES,
0830-0930
I
0~ 1
0
1
a
2
2
a
3
4
b) ILS-DME APPROACHES,
Figure 4.8:
I
a
5
6
7
Transmissions/20 second interval
1815-1915
Communications Transmission Histogram
-
84
-
2.
Transmissions divert attention away
from
other tasks as the pilot mon-
itors the message,
whether the mes-
sage is
to him,
addressed
aircraft,
3.
another
or the controller.
A message may invoke tensions, confusion or
anxiety in
the crewmem-
bers.
Because the
crew
occurrence of
workload it
is
voice communications
worthwhile
discussing some
features of ATC communications.
First,
voice messages is a stochastic process.
affects
important
the occurrence of
Messages occur with
some probability and associated with this probabilistic distribution is
ure 4.8
for
is
a mean message arrival rate and variance.
an example
of a typical
message occurrences
transmission histogram
probability
(i.e.
occurring in a given time interval).
Fig-
of n
messages
It possesses a charac-
teristic Poisson shape.
Because messages
mean message rate is
that
is useful
It
mechanism.
It
important.
shows
as that of
use,
is
though,
a set of transmission
loading
that probablility
figure 4.8 vary
85 -
the
one system parameter
and hour of the day.
-
manner,
the communications
should be noted,
approach in
for example,
a probabilistic
in quantifying
distributions such
weather,
occur in
by airport,
Figure 4.9,
histograms for a
'busy'
arrival into Los Angeles.
air traffic control
The stochastic nature of
messages is
quite complex
and deserves
special attention in pilot workload analysis.
discussion on this
contains a more detailed
of this thesis
The appendix
subject.
characterization
useful
of the
Boston arrival discussed previously.
as a
mechanism.
communications
are examples of
and 4.10b
Figure 4.10a
be constructed
can also
'message profile'
An ATC
this for
the same
Figure 4.10a displays
the number of messages sampled in an 80 second interval verFigure 4.10b, on the
sus the time into the flight mission.
utilization versus time into
indicates channel
other hand,
mission for this same message data.
It
is
the 'communications
figure 4.10a,
about the
to gain additional information
mentary manner,
nature of
in a comple-
to use figure 4.10a and 4.10b
useful
profile.'
For
example,
in
the number of messages/second during the vec-
toring portion
of the arrival is
rate during final approach.
tions (see figure 4.10b)
nearly the same
approach is that
about 50% of
However,
the message
the channel utiliza-
for these two arrival segments are
What is
(30-33%).
happening
tower control messages are
but of short mean duration.
during final
more numerous,
A typical Tower dialogue might
appear as below:
Controller:
Delta 11,
-
86
clear for take off
-
U
>-.-TI
C
L
0
14
No. of Messages/20 second interval
a)Center Sector
1|
2
3
4
5 6 7 8 9
No. of Messages/20 second interval
b)Approach Control Sector
.3'
.2.- r
1
..K1.F
2
3
4
5 6 7 8 9
No. of Messages/20 second interval
c)Tower Control Sector
Figure 4.9:
Transmission Histograms for a 'Busy'
LAX Arrival
-
87
-
Visual
Fl NAL -4
DESCENT
Time into Mission(sec.)
a) Smoothed Communications Message Rate
RMINAL
DESCENT
100
ijV
1000
500
b) Communi cations Channel
Figure 4.10:
TO0R IN G -
1500
into Mission(sec.)
Ti
Ut i zamtie
A Communications Profile
-
88
F1 NAL
Delta 11:
Delta 11 is
Delta 409,
Controller:
Delta 409:
This is
contrasted
rolling.
Position and hold 4R.
Position and hold 4R, Delta 409.
to the terminal vectoring
segment where
there are fewer messages/second but these, on average,
tend
to be longer in duration.
The communications
histogram,
time profile,
varies with day,
tion frequency.
weather,,.
Figure 4.11,
ent the profile is for
for data collected
like
the transmission
airport and communica-
for example,
shows how differ-
a Boston departure control frequency
during the same time period
as that for
The difference between figure 4.10 and 4.11 is
figure 4.10.
quite striking
and reiterates the importance
of communica-
tions in workload analysis.
Summarizing,
load the pilot.
air traffic control
communications tend to
They are characterized by a stochastic pro-
cess which interruprts,
commands and distracts the crew.
-workload studies it
therefore important to
is
It
magnitude of this loading source.
message occurrence rate,
sion time,
and averaged
averaged
is
In
identify the
suggested that mean
message rate versus mis-
channel utilization versus mission
time are system parameters useful in describing this mechanism.
-
89
-
C
0
U
0)
'U
U)
100
200
300
400
Time into Mission(sec.)
a) Smoothed Communications Message Rate
100
200
300
400
Time into Mission(sec.)
b) Communications Channel Utilization
Figure 4.11:
Communications Profile for a Boston
Departure Frequency
- on -
Chapter V
WORKLOAD ANALYSIS 0? A BOSTON ARRIVAL
Up to this
the air
the critical elements
point,
transport environment have
length.
To
comprehend the
these elements it
is
arrival of
been discussed
wide reaching
useful to examine,
a Boeing 707
in
at some
implications of
in detail,
This chapter,
ple of workload analysis.
of workload
an exam-
examines the
then,
aircraft into Boston's
Logan air-
port, breaking the arrival down into analytic detail.
Three forms of
These are
analysis are considered in
cockpit activity timeline
dence analysis,
communications.
task-prece-
Each of these approaches provides slightly
Each method
crew workload during the
yields evidence as to
the source of
as to whether the workload level
pilot loads and, a priori,
is
analysis,
of the air traffic control
and an analysis
different information pertinent to
arrival.
this chapter.
relatively higher or lower than some baseline level.
particular attention is
the analysis of this chapter,
to the influence
In
given
of the.air traffic control
system on crew
is
the basis for a
workload.
The scenario examined in this chapter
series of flight simulator experiments that were designed to
-
91
-
evaluate candidate measures of air transport pilot workload.
necessary to have a theoretical basis for
however,
It was,
comparison in order to understand the results of the experiments
then,
The analysis that follows,
results.
validity of
on the
comment
order to
and in
their
serves a dual
role:
1.
will
It
the
illustrate
concepts
outlined in the first chapters.
2.
It
a reference
serve as
will
for
the simulator experiments described
in later chapters.
THZ SCENARIO
5.1
The scenario
as Clipper
L4)
aircraft
involves a Boeing 707
(identified
that begins its arrival into Boston at 36000
feet over the Albany,
Clipper 54 is given a
New York VOE.
series of clearances that determine its arrival profile into
Boston.
The arrival
approach
point
after
terminates at the Catefory
has
the airplane
approach to Logan airport's runway
4R.
executed
II missed
an
ILS
Figure 5.1 depicts
the scenario graphically.
It
is convenient to segment the Boston arrival into three
portions.
below
The limits of the three segments are as indicated
with separate
for
references
timeline and the task precedence map.
-
92
-
the cockpit
activity
Albany,
VOR
NY
MANJO
Intersection
Gardner
VOR
118 deg.
magnetic course
Boston
Loaan
160
035
MILT
LOM
Plan View (Courses shown are magnetic)
a)
36000 ft
-j
22000
'4-
M)
6000
3000
00 ft
b)
Altitude Profile
A Boston Arrival
Figure 5.1:
-
93
-
-N-
Segment
Timeline
Precedence Event
Cruise Descent
0-840 seconds
A1-A3,B1-B5,C1-C2,
D1-D7,E1-E3
A4-A6, B6-B10,C3-C9,
840-1360
Vectoring
D8-D23
A7-B11-B13,C1O,
1360-1600
Final.
D24,E4
This segmentation is
segments has quite a
each of these
required cockpit
segment
performed because,
activity.
as will
distinct composition of
is hypothesized
It
be shown,
will have a significantly different
that each
level of average
workload than the other two.
5,2
ANALYSIS TOOLS
The following analysis makes use
of three analytic tools
each of which yields different information pertinent to crew
workload during the Boston arrival.
These tools are,
pectively,
1.
Cockpit Activity Timelines'
2.
Task Precedence Maps
-
94
-
res-
ATC Communications Analysis
3.
Because the
figures on which
these analyses are
based are
rather long and complex, they have been placed in Appendix B
of this thesis so as not to break up the body of the text of
the chapter.
figures in
Each page of the
panel number of the form P-XX.
Appendix B has a
The panel numbers are used
to refer to
particular sections of the
dence map.
Data
timeline and prece-
used in constructing the
chapter were gathered on numerous
figures in this
flight simulations of the
scenario discussed above.
5.2.1
Cockpit Activity Timelines
Task timelines are a historically useful tool for quantifying cockpit activity,
A timeline is essentially
a graph
that identifies cockpit tasks, gives estimates of task duration, and tabulates this information in chronological order.
Although task timelines have
they have some limitations which
tool in workload analysis,
should be noted.
First,
able or measurable.
not all cockpit tasks are identifi-
Second,
most mental planning tasks
For example,
therefore be omitted from the
cannot be identified and must
timeline.
been an historically useful
of arbitrariness exists in
some degree
the detail with which tasks are broken down.
in microscopic detail as a
a turn maneuver can be described
sequential combination of
many hand.,
-
95
For instance,
-
eye,
and
foot move-
ments,
each of which require fractions
cute.
Alternatively,
as a 'turn'
The
of a second to exe-
this maneuver can simply be described
of seven seconds total duration.
precision
of timelines
is
another
The time required
Appendix A.2).
vary between pilots,
limitation(see
to perform a
task will
vary with the conditions of flight, and
vary with the existing workload level.
Timelines,
best suited to describing only a nominal flight.
then, are
There will
be wide variations in practice from this nominal timeline.
The usefulness of task tirelines,
tations, should not be overlooked.
description of
Timelines can provide a
cockpit activity to
detail and accuracy.
It
is
despite the above limi-
a reasonable
also possible
degree of
to derive rough
estimates of physical effort requirements from the timeline.
Moreover,
timelines can readily be
expanded to reflect the
following:
1.
The
uncertainty
associated
with
each task's time of occurrence.
2.
The simultaneity of
multiple tasks
in the cockpit.
3. The
random
interruption
associated with
ATC and
communications.
-
96
-
process
intercrew
4.
identifiable plan-
The presence of
ning tasks.
The timeline
this thesis
format used in
includes enhance-
ments that are designed to highlight the workload parameters
mentioned above.
used,
5,2. 2
These features are explained, as they are
later sections of this chapter.
in
Task Precedence
aps
A task precedence map is another
be
cockpit activity
constructed from
represents a map
type of graph which can
timeline data.
It
events that occur
or schedule of discrete
in the cockpit where an event is
1.
The beginning or end of some physicockpit task(e.g.
cal
a
descent
maneuver.
2.
The occurrence of
a physical event
(e.g. the arrival of an ATC message
or the passage of a VOR)
The task
precedence
map
displays events,
uncertainty of these event times,
time of
event,
and the precedence const-
raints that must be maintained between events.
Task precedence maps differ greatly from cockpit activity
timelines both in
their construction, and in
task precedence map is
most useful
-
97
,
in
their use.
a very macroscopic
The
or
aggregate form where only major events are displayed.
identifying the way in
are useful in
other events.
events constrain
They
which certain cockpit
There exists
an unwritten
system of task priority among instrument pilots.
The prece-
dence map maps out these priorities and shows how the timing
series of tasks
of a
and execnwtion
requirement to execute some other series of tasks.
map displays
precedence
events,
constraints,
as
this hierarchy
a schedule
on the other hand,
which they convey.
The
timeline can be
of cockpit
activity, interruptions, busyness, and physical effort.
will be
shown later
that the
timeline and
actually complement one another in
Events ,on a
ated.
There is
event and
the precedence
are
denoted by boxes
along paths with which
they are associ-
physical significance to an
generally some
the path
to which it has been
aggregate event paths have been
assigned.
Position Event Path
(the geographic
position of the aircraft)
2.
Altitude Event Path
-
98
Five
defined for the purposes of
this thesis:
1.
It
workload analysis.
task precedence map,
and are positioned
of
are most useful in
analyzing the microscopic detail
referred to in
The task
and event flexibility or slack.
Activity timelines,
the minute detail
by the
constrained
is
-
3.
Communications Event Path
4.
Manual Control Event Path
5.
Planning Event Path
Event paths may be independent, physically connected or constrained in temporal order only.
An
event on one path may
be constrained to precede an event on another path.
denoted by a
A constraint is
solid line between any two
solid or a dotted line.
events indicates that a physical
task occurs between the two events and the duration,
onds,
of this task is
A
in
sec-
indicated by the number in parentheses
located on the constraint,
approximately midway between the
two events(see for example figure 5.2,events B1 and B2).
A 'soft' precedence
constraint is indicated by
line connecting two events
two events
constraint,
map
It is
irplied,
that are
then,
connected by
to precede
the
that the time beta soft
precedence
must be greater than or equal to zero seconds.
The implication
dence
is constrained
the left
event to
event on the right.
ween
(see events C1-D1 of figure 5.2).
being performed between two such events,
No physical task is
but the
a dotted
displays
'earliest' and
of all the
constraints that
graphically
a 'latest'
is
that
there
at which each
tiine*
the preceis
an
event can
*Event times on the precedence map are taken to be deterministic. With some simplifying assumptions (i.e. norral-
99
F.
occur.
These limit times appear, in seconds, as the number3
in parentheses beneath each event box.
'latest' time is the rightmost
the
is the leftirost number,
The 'earliest' time
-
number.
The differencebetween the earliest
referred to as the slack.
an event is
and latest times for
strictly
The slack,
speaking, is a measure of the uncertainty in the event time.
It
is
precedence
flexibility a
tasks associated
Slack is that flexibility allowed the crew-
with an event.
by all
as the
executing the
crewmember has in
pilot or
member
thought of
more pragmatically
the
constraints shown
interacting
on
the
The existence of slack can be depicted on a
map.
timeline by allowing
tasks to slide from
their nominal (or
earliest) time to a Task Limit defined by the slack value of
a precedence map.
Given some degree of task flexibility an operator working
on multiple tasks is
manner
as to
likely to perform these tasks in
reward
minimize some
variable (see
such a
Tulga).
Instrument pilots are trained to perform tasks on a priority
basis that uaximizes safety and that minimizes workload ( or
task demand
rate).
If
no slack
exists for
a particular
event, then its associated tasks must be worked on promptly,
increasing task demand rate and thus instantaneous workload.
A path
along which the slack
is always zero is
ity) stochastic event times can also be used.
-
100 -
called the
Posi ti on
Events
A
-
A2
Cross
VOR B
-(180)
Cross
Sector
A
A
Al t i tude
Even ts
...-..10..-
\nLeave
10000ft
18ooft
Communi cations
Events
Cl
Reduce
to 310
knots
\by
--
Begin
Slowing
Maneuvr
E
El
E2
-
Begin
nnc
C
D2
BegD
Slowing
250 kts
D1
D
Planning
Events
C2
Turblnce
Reportet
707
(160)
Manual Control Events
E
B
Leave
100
0
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-E
Turblnc
Planin
nni
Figure 5.2:
A Task Precedence Map
i
'critical
path'
double line
and is
denoted on
the precedence
map
by a
(see Path A of figure 5.2).
the task precedence map is
_o summarize,
load analysis because of its
dence constraints,
capability to highlight prece-
task priorities,
The precedence map,
useful in work-
though- derived
and task flexibility.
from timeline data,
is
much more macroscopic in its scope and in its use.
Analysis of ATC ajnd Intercrew commun 1ations
5.2.3
Section
four of
chapter four
discusses the
through which ATC communications may
In a
similar manner
mechanisms
tend to load the-crew.
intercrew communications
may tend
to
interrupt and distract the crew and thus induce higher workload levels.
of voice
Due to such effects the statistical properties
communications messages will
Boston arrival.
be examined
Particular attention will be devoted to
1.
Interruptions due to messages
2.
Tasks triggered by messages
3.
Effects
(.i.e.
of
Channel
Utilization
Busy Air Time/Total Mission
Time) and message rate
4.
for the
Comparison
of
message
between mission segments
-
102
-
properties
5.3
AN ANALYSIS
In this section,
six parameters are examined that (by the
cate relative levels of pilot workload.
1.
and that indi-
characterize the nature,
theory of Chapter 2)
Busyness -
These are:
the amount of time
i.e.
cockpit tasks and
spent working on
only moni-
or performing
not idle
toring tasks.
2.
Task Multiplicity and Interruptions
-
degree
The
to
multiple
which
tasks are present and the magnitude
of
task interruptions
which
then
results.
3.
Task Simultaneity
degree to
- The
which multiple tasks occur simultaneously.
4.
Task priority
the priority structure
ing different
Examine
-
and Slack
for execut-
tasks and
determine
the time flexibility or deferrability
which
each
task
is
then
allowed.
5.
Constraints
ority
and Triggers -
structure defines
-
103 -
The pri-
a set
of
constraints between
constraints identify
of the ATC/cockpit
These
tasks.
-
whichelements
system strongly
influence, crew activity
and
also
identify which events trigger other
events to occur.
6.
ATC-and intercrew
The
conmunications -
statitistical
messages and the
properties
of
activity messages
invoke.
Each of
the above
parameters are
three scenario segments (cruise descent,
and final aFproach).
investigated for
the
terminal vectoring,
The section that follows
contains a
concluding synopsis of the analysis.
5.3.1
Busyness
Chapter two categorized tasks
into three broad classifi-
cations:
Operating tasks ately,
Tasks which must be
such as responding to
an ATC
handled immedimessage or cont-
rolling the aircraft.
Monitoring time period,
Tasks which can
e.g.
be deferred for
systems monitoring.
-
104 -
a small
Table 5,la) Number of Tasks for the Boston Arrival
tt t . --
'
(Excluding all "Hold" category and communications tasks.)
Headitn
AltltuAe
Speed
Nav /PIann ng
Total
Cruise
Descent
7
18
3
2
30
Terminal
Vectoring
9
16
4
11
40
Final
Approach
4
6
3
8
21
Table 5.1b) Busy Time (seconds)
(Excluding all "Hold" category and communications tasks.)
Total
Segment
Heading
Altitude
Cruise
Descent
820
121
Terminal
Vectoring
174
104
42
75
309/520
Final
Approach
246
173
32
38
233/240
Speed
Nav/Planning
840/840
*
Note:
Total is not the summation in this case, since many tasks
occur simultaneously.
-
105
-
Table 5.2: Mean Task Interarrival Time (seconds)
Nav/Planning
Mission
Segment
Heading
Altitude
Speed
Cruise
Descent
1#
43
390
550#
31
Terminal
Vectoring
86
32
97
46
17
Final
Approach
13
16
33
35
16
*
Notes:- * Includes all tasks except communications tasks.
# indicates datum unreliable due to small sample size.
Table 5.3: Mission Mean Task Interarrival Times (seconds)
Including Communications
Tasks
Excluding Communications
Tasks
Segment
All Tasks1
All Tasks3
All Tasks for 2
Pilot Flying
All Tasks for 4
Pilot Flying.
Cruise
Descent
31
31
31
31
Terminal
Vectoring
17
17
13
15
Final
Approach
16
18
12
12
Notes:
1 - Excludes communications tasks.
2 - As specified by flight manual, also excludes comm. tasks.
3 - Includes intercrew communications and ATC messages
to/from CL 54.
4 - Same as above, but includes only intercrew messages
which pilot must command or respond.
-
1O6
-
Planning -
Tasks which
can be
deferred for
a longer
such as approach planning.
period,
an IFE,
The nature of the piloting tasks in
environment is
air transport
such that there are phases of flight when the
crew is nearly idle (performing mostly monitoring tasks) and
there are phases of flight when the crew is
three categories of task).
(working on all
is
extremely active
Busyness,
then,
one parameter which characterizes the workload level of a
It
crew.
task load (in
extent,
be quantified to some
may
terms of number of tasks
(an indication of relative difficulty),
the percentage of segment time
on
working
at
least
one
by measuring
active),
task type
and busy time
(i.e.
during which a crewmember is
task).
The Cruise Descent Segment
At fourteen minutes in duration,
ment is the longest in
that
this segment
includes a
steps at 22000, 14000,
changes
the most
It
long
and 8000 feet.
Figure 5.1 indicates
descent with
altitude
There are two course
The segment length seems to be
during the descent.
notable feature of
this portion of
the scenario.
is an important feature because,
if workload is measured
14 minutes,
the instantaneous peaks
as an average over the
in
the scenario.
the cruise descent seg-
workload will be smoothed out by the averaging process.
-
107 -
Panels 1-9 of figure B.1 contain the task information for
the tracking task associated
panels shows that
the primary task and that it
to a VCR is
other tasks generally
verifies this
Panel 4,
than 20 sec-
the data of figure B.1
by itself,
for example,
Excluding
VOR tracking task as
there are 30 tasks occurring in
The number of tasks is,
these
and that
observation quantitatively.
communications tasks (and counting the
one task),
with flying
also clear that there
have a duration of less
Table 5.1a was compiled from
onds.
of these
persists throughout
occurring
other tasks
relatively few
and
is
From the timeline it
the segment.
are
A brief examination
segment.
the cruise descent
840 seconds.
misleading.
Vieving
some tasks are continuous (e.g.
while other tasks
task),
heading category tracking
the
are of
rapid succession
with other
tasks (see tasks 6,7,8 in the altitude category).
Thus, the
short duration,
small number
of tasks does
crew is idle.
is
ing
but occur in
not necessarily imply
From Table 5.2 it can be seen that the pilot
busy working on at least one task,
the cruise
that the
Moreover,
descent.
include hold category
tasks.
100% of the time durTable
5.2 does
These tasks require intermitpilot to some extent,
tent crew attention and load the
are not included in the data of Table 5.1 or 5.2.
he Te minal Vectoring Segment
-
not
108
-
but
The
terminal vectoring
minutes in
segment
is approximately
duration or slightly more
duration as- the descent segment.
than half as long,
in
Panels 9-14 of figure B.1
describe the cockpit activity for the segment.
parison of these panels with Panels
toring segment is
eight
A cross com-
1-9 shows that the vec-
much more complex,
much
busier than the
descent seguent.
Figure 5.1 shows that the
ario is
vectoring portion of the scen-
dominated by four turns,
turn to 130,
a turn to 160 degrees,
and a turn to
a turn to 080,
final approach
course.
three required
altitude changes
Moreover,
are descents from 8000 to 6000 feet,
and from 3000 to 2000 feet.
intercept the
this segment
(see figure
a
5.1b).
includes
These
from 6000 to 3000.feet,
This aircraft maneuvering seems
to dominate the activity on the timeline of Panels 9 through
Table
14.
5.1a indicates
altitude category
that there
tasks for the
are 25
heading and
vectoring segment
as com-
pared to 25 heading and altitude category tasks for the descent segment.
are squeezed,
However,
now,
these 25 aircraft maneuvering tasks
into half the
tine frame of the cruise
descent.
Navigation and planning tasks
(28')
of the 39 total
segment.
This is
planning tasks in
compose a major percentage
tasks occurring during the vectoring
in comparison with only two navigation and
the descent segment.
-
109
-
Six of the 11 plan-
ning tasks occur in rapid sequence
the performance of the 'before
line and are associated with
landing'
Again,
on Panel 10 of the time-
checklist.
Table 5.2 indicates that (assuming one crewmember
leading.
the pilot is
tasks)
performs all
task 60% (309/520)
least one
composition of tasks
however,
indicate
is much different
than that
There is
for the descent segment.
cruise descent
time for the
the timeline,
Panels 9-14 of
segment,
working on at
time during the descent segment.
of the
This is coirpared to 100% busy
that the
is somewhat mis-
by itself,
the number of tasks,
a greater variety in the
number and type of tasks than those which present themselves
in
Panels 1-8.
the vectoring
occur during
Tasks which
portion of
the
mission are more numerous, but of shorter mean duration than
those during the
timeline it
is
From observation
cruise descent.
obvious which type of situa-
not immediately
tion tends to induce higher workload levels.
situation,
This
is
short dura-
long duration tasks,
rather than few,
higher level.
The theory of
case of many,
Chapter 2 would suggest that the
tion tasks,
of the
due to the fact that
in
induces a
the former
tasks are continually intercupted or preempted by
theory suggests that,
Thus,
tasks of higher priority.
dur-
induced levels of workload may be
ing the vectoring segment,
higher than during the cruise
-
110
descent segment,
-
despite the
fact that the busy time for
this segment is
than that for
the vectoring segment)
time for
lower
(60% busy
the cruise
(100% busy time).
descent
Final Approach Segment
The final approach is characterized
stant descent-rate
maneuver.
mission segment is
four minutes,
vectoring segment.
The
duration of
con-
this final
duration of the
half the
Examination of
the 'track localizer' and
by a straight,
Panels 14-16 shows that
'track glideslope' tasks dominate
the activity of this segment.
As indicated
which must
in Table
5.1a,
there
are 21
the segment..
be performed during
total tasks
This
can be
compared with 40 tasks for the vectoring segment which has a
duration that is twice that
of this final approach segment.
The composition of these tasks appears similar to the composition of tasks for the vectoring segment.
tasks
(6/21)
Again 28% of the
are navigation and planning tasks.
The number of tasks should not be used alone in quantifying cockpit activity on final approach.
the two simultaneous track-
'Number of Tasks' metric weighs
('track
ing tasks
with other,
localizer'
This is because the
and 'track
glideslope')
(e.g. Seat Belt Sign - On).
less difficult tasks
The busy times of Table 5.1b complement the 'Number
data by
the time the crew
indicating how much of
-
111
equally
-
of Task'
is busy,
regardless of the number of tasks.
The table shows that, on
final approach, the crew is working on at least one task 95%
(233/240)
of the segment time.
Further examination of
Table
5.1
final
approach segment
show
Panels 14-16 of the
that the
composition of
is
a
many tasks
tasks during
combination of
cruise descent and vectoring segments.
(21 in 240 seconds)
timeline and
the
that for
the
There appears to be
some of which are long dura-
tion (the tracking tasks)
and some of which are short dura-
tion
Planning tasks).
(the
Navigation and
Based on
the
busyness of this segment one would expect the measured work-
load level to te approximately the same or greater than that
for the vectoring, but greater than that for the cruise descent segment.
5.3.2
Aultiple Tasks and Interruptions
The Cruise Descent Segment
At any
particular point in
the activity timeline
evident that the piloting task involves multiple,
ing tasks.
it
is
interrupt-
The timeline format of figure B.1 highlights the
occurrence of this interruption
process.
Panel 7 provides
an example of this feature.
In the time period 620 to 680 seconds,
a course change,
a level off
-
the pilot pertorms
followed by a further descent
112 -
initiation
(see Alt.
navigation course
scanning the
colulmn)
( see Nav/Planning
setting
Panel horizontally
time (660 seconds)
along a
column) .
line of
By
constant
the multiplicity of tasks becomes appar-
At 620 seconds a speed
ent.
and he also must reselect his
and and altitude hold*
task
are being monitored while the heading category tracking task
is
being
performed.
At 670 seconds
a left turn
executed simultaneously with the execution
is being
of a short dura-
tion, three task altitude sequence.
Quantifying the
magnitude of
the interruption
Table 5.2 gives mean task interarrival times.
average time between
the occurrence of two
the timeline in figure B.1).
small (refer to
some of
It does appear,
most to
though,
(based on
are unreliable.
that the altitude tasks contribute
the interruption process
the scenario
tasks
cruise descent segment)
interarrival times
rate of 1 task every 43 seconds.
on
That is, the
Because the number of tasks is
Table 5.1a for the
these computed
process,
with a
mean interarrival
This is as expected based
structure portrayed
in
figures 5.1a
and
5.1b.
are
holding a descent rate,
e.g.
*Eold category tasks,
that run
indicated by the solid lines with arrowheads,
through the center of task columns on the timeline.
*Mission mean interarrival time is computed by grouping
maintaining the chronological
all tasks in one category,
mean interarrival times.
computing
relationships, and then
one task.
as
Simultaneous tasks are counted
-
113 -
The Mission*
mean interarrival
emphasizes how multiple,
In Table 5.2,
however,
is
which is
25X
better
interrupting tasks load the pilot.
the Mission interarrival time
between tasks
tasks,
time,
smaller than
the 43
of 31 seconds
seconds between
indicated for altitude tasks.
In this long
descent segment it is the altitude tasks which are most numerous (17 altitude tasks, see Table 5.1).
dom
occurrence
of
tasks in
three
However, the ran-
other
categories
has
reduced the Mission mean task interarrival time from approximately 43 to
arrival
31 seconds between tasks.
rate,
induced
by the
This higher task
multiple task
environment,
should tend to increase the workload level of the crew above
that encountered in a single task mode.
Terminal Vectoring Segment.
Panels
9 through
14 of
the
indicate that the multiple task
ated interruption process
is
cockpit activity
timeline
environment and its
associ-
much more evident
in
toring segment tnat in the cruise descent segment.
across the timeline
this more apparent.
of heading,
at 940,
At these
altitude,
occurring either
1240,
and
manner.
-
114
a combination
and Nav/?ianning
simultaneously or in a
-
Scanning
1320 seconds makes
mission times
speed,
the vec-
rapid,
tasks are
sequtntial
Table 5.2
task interarrival
times based
on
The altitude and the Nav/Planning task catego-
figure B.1.
ries have
gives mean
the smallest task
quent task occurrence)
interarrival times
at 32
(most fre-
and 46 seconds between tasks,
respectively.
For the altitude category tasks this is
ler
43
than the
second figure
cruise descent segment.
which
smal-
obtains during
the
Heading tasks, with mean interarri-
val times of 86 seconds would seem to be the least demanding
task category
with respect to task
interruptions,
despite
the fact that the vectoring turns are a predominant physical
task during this segment.
The dission
mean interarrival times
reflect the marked
shown in
impact of the multiple
task environment
on crew task load during the vectoring segment.
mean interarrival time of 17 seconds
Table 5.2
The Mission
is 50% of the interar-
rival times for altitude tasks alone and 20% of the interarrival time for heading tasks alone.
four
task categories
The interaction of the
(communications tasks
have not
counted)
in quadruplex has the effect of dramatically
ing the
task interarrival time
perceived by
The 17 second task interarrival
ment is
been
reduc-
a crewmember.
time for the vectoring seg-
roughly half that for the cruise descent.
The crew-
members should thus perceive themselves as being much busier
and find tasks interrupting one another more frequently during the vectoring segment than the descent segment.
basis of perceived busyness due to interruptions,
-
115
-
On the
one would
for the
relatively higher
to be
expect workload
terminal
vectoring segment of the Boston arrival.
Final Approach Segment
Compariscn of
the final approach
the degree to
tions caused by task
which multiple tasks
and distrac-
Scanning across
interruptions occur.
-at times 1380
figure B.1
terminal vectoring
that they are roughly equi-
segments on figure B.1 suggests
valent in
and
and 1460 seconds,
for instance,
in most task categories sim-
shows that tasks are occurring
Beyond time 1500 seconds there are two simul-
ultaneously.
taneous tracking tasks and one 'hold'
task being performed
category speed control
by the pilot as he
flies the aircraft
down the Instrument Landing System.
Table 5.3
indicates that
altitude category
(ignoring the heading cate-
most often during final approach
gory due
very
segment.
same as the
close to the
seconds between
at 16
value for the
also 16 sec-
terminal vectoring
Because the Mission mean interarrival time is the
mean interarrival time of
data suggest that
as in
size)
Mission mean task interarrival time is
tasks.
onds,
to the small sample
tasks occur
altitude tasks,
task interruptions are not
the terminal vectoring segment.*
the
as pronounced
However,
another
*In the terminal vectoring segment the Mission mean task
least 50% of the value of the
interarrival time was at
time for all task categories
minimum task interarrival
task
This effective reduction in
table 5.2.
shown in
116
-
effect is taking place that is
Table 5.3.
not reflected in the data of
This is the effect of task simultaneity which is
discussed in a later section.
Task Simultaneity
5.3.3
The mean
the
fact
task interarrival times discussed
that
instance,
some
tasks
occur
on Panel 7 of figure
left turn to
simultaneously.
B.1,
111 degrees and a
above ignore
it is apparent that a
descent initiation sequence
occur simultaneously at 667 seconds in to the scenario.
presence of task simultaneity will
but
is not
metric.
measured
well by
For
The
thus affect crew loading
the
task interarrival
time
This section examines the degree to which task sim-
ultaneity is
p-resent in
each of the three arrival segments.
The Cruise Descent Segment
The degree of task simultaneity during the cruise descent
can be inferred from the Task Simultaneity histogram of figure 5.3.
In
this figure the number of occurrences of simul-
taneous tasks have been counted based on the nominal time of
task occurrence shown in figure
B.1.
These data were then
normalized to construct the histogram.
interarrival time
was said to be due to an interruption
process caused by the presence of multiple tasks.
-
117
-
.9
.8
.7
.6
.5
.4
pp
.3
.2
.1
J
~
1
2
2
3
3
B
I
4a
4
5
5
Simultaneous Tasks
a)Cruise Descent Segment-Excluding Communications Tasks
1
2
4
5
Simultaneous Tasks
3
b)Cruise Descent Segment-Including Communications Tasks
Figure 5.3:
Task Simultaneity Histogram for the
Cruise Descent Segment
-
118
-
1
2
3
4
5
Simul taneous Tasks
a)Vectoring Segment-Excluding Communications Tasks
1
2
3
4
5
Simul taneous Tasks
b)Vectoring Segment-including Communications Tasks
Figure 5.4:
Task Simultaneity Histogram for the
Terminal Vectoring Segment
.9
.8
.7
.6
.5
.4
.3
.2
1
2
3
4
5
Simultaneous Tasks
a)Final Approach Segment-Excluding Communications Tasks
1
2
3
4
5
Simultaneous Tasks
b)Final Approach Segment-including Communications Tasks
Figure 5.5:
Task Simultaneity Histogram for the
Final Approach Segment
-
120
-
Figure 5.3a indicates that there are frequent occurrences
of
heading tracking
the
indicates
that
throughout
the cruise descent.
any
other category
this tracking
so task
is
timeline
continuous
a task occurring in
performed simultaneously
Nevertheless,
occurrences of three,
task
Thus,
must be
task.
of the
Observation
tasks.
two simultaneous
with
shows few
figure 5.3a
or five simultaneous tasks and
four,
siultaneity should
not significantly
affect crew
busyness during the cruise descent.
The Terminal Vectoring Segment
The
teriinal
Figure 5.4a shows that there is
tasks occur.
bility that two
ously.
seen in
that three tasks will
comparing figures 5.3a
the vectoring segment
task).
This
tasks
are more
and 5.4a.
there is a high
occur simultane-
That is,
consistent with
is perfectly
during
probability that one
(not simultaneously
by itself
and a much
these segments can be
The major difference between
task will occur
a large proba-
tasks will occur simultaneously
smaller probability
the
which simultaneous
the degree to
in
cruise descent segment
similar to
segment appears
vectoring
with another
the fact
that
mean task
interarrival
times are smaller during the vectoring segment.
Tasks occur
more often,
numerous and
that
but they simply do not occur simultaneously.
The Final Approach Segment
-
121
-
Task
the earlier mission segments.
during final approach than in
This is
that the
5.5a which indicates
evident from figure
three,
frequency of occurrence of two,
tasks is
consideration
important
more
a
is
Siirultaneity
and four simultaneous
significant.
terminal vectoring
and the
approach
5.3.2 showed
of section
The discussion
multiple tasks occur-
a high incidence of
'final'
there is
ring,
which the
task
interarrival
the incidence
occurring simultaneously
shows that during
however,
Figure 5.5a,
nearly
time metric
on final
approach,
ignores.
four tasks
and
three,
of two,
the same for
another is
degree to which tasks interrupt one
Because of
segments have
final
This suggests that the
identical task interarrival times.
these two segments.
that the
the
busyness
perceived by the pilot may be greater on final approach than
Due to the joint influence of
during the vectoring segment.
total workload
uncertain whether measured
is
it
however,
other parameters,
than during
during 'final'
will be greater
the vectoring segment.
5,3.4
.T'ask Priority and Slack
In section 5.1.1,
it
timelines
are inaccurate
associated
with the
timeline.
That is
was explained that cockpit activity
because there
uncertainty
is an
nominal task execution time
figure
(see Panel 7,
to a heading of 111 degrees
-
B.1),
shown on a
the left turn
may be initiated at 627 seconds
122
-
or it
may be initiated later
task to
be executed.
The current
than the
more important
workload level
task initiation time.
pilot can also affect the
may elect
if there is a
to begin work on
nominal task time
The pilot
a task either earlier
shown on an
of the
or later
activity timeline,
this decision being based on what his experience,
training,
and workload level indicate to him.
The task precedence
map discussed in
quely determines the limits within
These task limits
parentheses,
which a task must occur.
are indicated by the
numbers enclosed in
The difference between the second number of the
pair and the first
number is
equivalent to the total
'earliest
5. 1.2 uni-
located beneath event boxes of the task prece-
dence map.
is
section
referred to as the 'slack'.
time range
It
(beyond the so-called
tine') through which the task may be slid.
Slack, then is an indication of the priority of a task or
of the
time flexibility
task.
The trained
that a
pilot has
IFR pilot would be
in
executing
a
expected to utilize
event slack in a manner which allows tasks to be executed in
a time sequence that minimizes his probability of error.
the timeline
of figure B.1,
nominal time
of occurrence
all
tasks are shown
(their 'earliest
time').
In
at their
Task
limits for a task are denoted by dark bars and an accompanying TL 1-9
(Task Limit for task 1 thhough 9 of this column).
The slack,
on the timeline,
-
is thus the difference between
123
-
the time
the TL bars
at which
are shown
and the
time at
section 5.1
for the
which the associated task occurs,
jhe Cruise Descent Segment
For the
long descent
limits of each
portion (see
mission segment on the
figure B.2 shows that the average,
large.
task precedence map)
noncritical slack is
On Panel 1 of the figure, no slack (off the critical
path) has a value of less than 100 seconds.
onds,
The minimum
which were shown
noncritical slack value of
again associated with
is
(see event C2,D3)
in table 5.1 to be the
type for this segment.
D4-D6,
What is
tasks.
Events
slack values
E1,E2,
and 940 seconds,
and E3
early in
associated with
show slack values
the descent,
but these tasks
In
of 1600,
will be
important operating
the air transport environ-
operating and monitoring tasks
while another executes planning tasks.
-
planning
As discussed in Chap-
are preempted by more
one pilot may execute
figure B.2
to start the execution of plan-
or monitoring category tasks.
ment,
1 and 2 of
respectively.
the pilot will tend
deferred as they
most numerous task
on the critical path (see
rost notable from Panels
very large
ning tasks
altitude tasks
and E3).
are the
ter 2,
166 seconds
Panel 2 of figure 3.2 also indicates
large slack values for events not
events B5,
Paths B, C, and
slack values greater than 300 sec-
E of the Panel have mean
1220,
very
124
-
the large slack values have
For the long descent segment
important implications:
1.
Though the
planning tasks
require
substantial time to execute
70
indicates a nominal
Panel 2,
B.1,
(figure
seconds for
review the approach procedure)
task flexibility is
that its
to
one crewmember
the
large enough so
presence is
not likely to
induce high workloads.
2.
Panels 1-9 of the timeline indicate
a few
occurrences of
Again,
tasks.
simultaneous
due to slack valu-es
that range from 166 to 1600 second,
the
occurrence
tasks is
high
also not
work
of
simultaneous
likely to induce
loads during
the
long
descent segment.
The Terminal Vectoring Segment
A brief examination of Panels 2-5 of figure B.2 indicates
that the
terminal vectoring segment*
descent.
plexities than the cruise
has many
Despite this,
more comfurther
events A4-A6,
AOn the task precedence map of figure B.2,
D8-D23, inclusive, contain the terminal
C3-C9,
B6-B10,
vectoring segment.
-
125
-
examination of the figure shows
that event slacks,
off the
are smaller than the slack values for cruise
critical path,
descent segment.
In
all cases
are less
the event slacks for
than 415 seconds.
this segment is
D18
The minimum nonzero
46 seconds which occurs for
(see Panel 5,
rences of
figure
slack values
slack in
events D17 and
There are numerous occur-
B.2).
less than
slack values for the segment
cal path)
the vectoring segment
120 seconds.
The mean
(excluding events on the. criti-
are listed below.
Event Path
Mean Slack(seconds)
A
246
B
171
C
0
D
133
the terminal vectoring segment
The mean event slacks for
are much
Most notable are
descent segment.
values for the
straller than the
the 171 and 133 second
on path B
and D,
values of
587 and 356 seconds
respectively.
-
126
values for
These can
be compared to
for the mean event
-
events
slack on
This sug-
paths 3 and D during the cruise descent segment.
gests that the
time in
crew will have much less
which they may
flexibility in
execute required tasks.
the
On average
tasks may only be slid 30% of the time distance allowed during the
intertask constraints will cause workload
ing the
than during
vectoring segment
that the
This indicates
cruise descent segment.
to be higher durdescent
the cruise
segment.
Final Approach Segment
8.2 describe the nature of the
Panels 6 and 7 of figure
event slacks for the final approach segment.
ble feature
that most
The most nota-
on the critical path
of the events lie
map is
the task precedence
of this portion of
and thus
have -slack values of zero seconds.
6 and 7 of the figure
Scanning across Panels
it can be
off the critical path (Events
seen that the only two events
A9 and C10)
have slack values of 123 and 213 seconds.
values are
comparable to the
vectoring segment and
the mean values),
average slack values
much smaller
cent segment.
127
for the
of 30% of
values for the cruise des-
than the mean
-
(on the order
These
-
tireline of figure B.1
The cockpit activity
little
more information
final approach
14-16
on the effects of
of figure
B.l shows
occurring and these
slack during the
the scenario.
portion of
that there
Scanning Panels
are numerous
are occurring in multiple
ries (refer to sections 5.2 and
provides a
task caiego-
Since most of these
5.3).
tasks are associated with events on the critical path,
ever,
promptly.
how-
flexibility and are constrained
they possess no time
to be executed
tasks
several tasks that nomi-
Moreover,
nally occur during the cruise
descent and vectoring portion
of the missions have slack values that allow their execution
to be
slid well
approach segment
into the final
14 and 16 of the figure).
example the Task Limits on Panels
This suggests that if
sually busy,
the terminal vectoring segment is
unu-
some tasks may be deferred for execution duLing
final approach.
Because additional tasks must then be exe-
cuted within
the increasingly tight constraints
the critical
path,
result.
(see for
This
a
higher task
effect is
imposed by
demand situation
caused by
the low
would
average slack
value during final approach.
then,
is much more tightly
constrained than the vectoring segment.
Most of the final
The final approach segment,
approach tasks
path and
are associated with
therefore must
afforded large
be executed
events on
promptly rather
degrees of time flexibility.
-
128
-
the critical
than
Whether this
tightly constrained
level for the
situation results in a
higher workload
final approach segment will be
whether perturbations
to cockpit
activity
a function of
(e.g.
an emer-
gency) during earlier flight segments cause tasks from these
earlier segments to be deferred into the tightly constrained
final approach segment.
5.3.5
Constraints and Triggers
In workload analysis it is
crew loading.
que in
useful to examine sources of
ihe task precedence map of figure B.2 is uni-
this context as it
ATC/cockpit system
indicates which elements
are constraining
one another
of the
and which
ATC/cockpit events trigger other to occur.
The Cruise rescent Segment
The discussion of section 5.2.2
(the double lined
tical path
which events are
path in figure B.1)
imposing the primary constraints
Panels 1
triggers.
explained that the cri-
and 2 of figure B.2 show
duration of the cruise descent segment,
That is,
the aircraft's position
gering most
sented in chapter 4,
the air
and task
that for the
?ath A is critical.
along the airway is
Following the
cockpit events.
indicates
trig-
arguments pre-
it is the airspace loading mechanism of
trafic control
system that
is
triggering
activity and inducing the predominan't cockpit load.
-
129
-
cockpit
present on Panel
B5 cues
2 of the figure where at
the occurrence of a
aircraft altered its
path of
in the
a change
airspeed,
250 knots
slowing maneuver to
is
the airway
triggering most cockpit activity.
structure that is
If
the
cruising
flight oc changed its
distribution of
time
also
10000 feet Event
path A suggests that it
However,
(Event D6).
is
the Regulatory Loading mechanism
The effect of
cockpit
activity would result.
due to
indicates some event triggering
Figure B.2 also
flight procedure.
In particular,
flight procedure calls for
two checklists to
be executed one at 18000 feet
and one at
This effect is denoted by the soft-precedence
10,003 feet.
constraint connecting Event E4 and E2 and that connecting B5
and 23.
Observation of the cockpit activity timeline in
vicinity of these events
indicates that
loading at
occurrence
limits
(see P5,P8-9,
these events
a busy time nor
do not
particular cockpit
enough (1220 and 940)
loreover,
task
induce the
the task
procedures are
large
so that high workloads are not likely
to be induced.
Terminal Vectoring
however,
greatly increase
does this procedure
of many simultaneous tasks.
for these
figure B.1),
the
Segment
-
130
-
figure B.2,
Observation of Panels 2-6,
the terminal vectoring segment the
for the major portion of
communications path.
the
along Path C,
critical path lies
indicates that
That is,
during the vectoring segment most events are trig-
gered by
air traffic control communications
the execution and
constrains
communications events
between these
the timing
Also
events.
are associated with
timing of tasks that
events on non-critical paths.
Boston arrival provides an
The vectoring portion of the
referred to in
example of
what was
loading of
the pilot due
to air traffic
control messages.
for instance, trig-
Event C4 (refer to Panel 2, figure B.2),
gers the occurrence of the Begin Descent event,
27,
28
category
in the altitude
(see Panel 12,
D8.
event are timeline
this Begin Descent
ated with
as explicit
chapter 4
category and
figure B.1).
task 5 in
Associtasks 26,
the speed
Panel 4 of figure B.2
shows how communication event C6 triggers a turning maneuver
associated
with events D13 and D14.
air traffic control
then,
In this segment,
messages exert a dominant
influence on
the nature and timing of the crew's task load.
the vectoring segment where
There are other portions of
tasks are triggered
example,
event A5,
by events off the
figure B.2)
-
triggers a turning event
1-31
For
at Manjo intersection
the aircraft's arrival
P2,
critical path.
-
(see
(Event
D10)
Associated
8,9 on
heading tasks
These tasks
figure B.l.
of
Panel 10,
D10 are
with event
are triggered only
indirectly by the critical path.
Final Approach Segment
The critical path for the final approach segment wanders
figure B.2).
cockpit
B (see P6-7,
During the early stages of the final approach,
position along
tasks are triggered by the aircraft's
An example of this is
the localizer.
event E4 on Panel 6 of
event
by the localizer interception
triggered
essen-
event is
Checklist'
This 'Eegin
the precedence map.
tially
and then to path
A to path D
from event path
(note
Panel 14
that event E4 and A6 have the same earliest time).
of the cockpit activity timeline indicates that Nav/Planning
tasks 14,15 and
associated with
communications task 43 are
this event.
As
the aircraft
arrives
at
the point
glideslope
of
becomes critical.
interception,
the altitude path (Path B)
The aircraft's
position on the glidepath then triggers cock-
pit events.
Figure 8.1
occurs.
provides the
for the extension of 40
glidepath.
Similarly,
-
132
degrees of flap
position in
are both triggered by the aircraft's
the
how this
for and selection of landing
On Panel 15 the call
gear extension and
detail on
communications
-
relation to
events
56-63
(Panel 16)
aircraft's altitude,
also triggered by the
are
once the glidepath has been intercepted.
It
clear from
entirely
airspace structure (i.e.
glidepath angle,
The
ing.
etc.)
structure of
workload and
again the
aircraft
operations manual,
loads.
approach speeds
the critical
and
paths.
as specified
in the
significant role
the
in
the specified
in particular,
If,
clearly
altitude event
plays a
were changed,
path would be affected.
this is
the location of the cri-
flight procedure,-
However,
inducing task
the
approach procedure
the instrument
the position
path along
map whether
the primary cause of pilot load-
is
indicated on the precedence map by
tical
not
the instrument approach procedure,
affect crew
definitely does
precedence
the task
is
final approach it
should be noted that during
length of
the critical
In some circumstances the length of
the nominal
path could affect
workload level
(see appendix A.2).
5.3.6
ATC and Intercrew Communications
suggests that the structure
The discussion of chapter 4
of air
traffic control
communications
to occur.
induce explicit
Similarly, required inter-
and implicit loads on the crew.
crew communications can cause
can
interruptions and distraction
these communications is
The nature of
important element of this workload analysis.
-
133
-
thus an
5.6
Figure
depicted in
was
the
using
constructed
column on
the communications
dialogue
of
the timeline
It represents a 'low' message rate relative to
figure B.1.
Figure 5.7 repre-
some- of the data presented in chapter 4.
sents a
ATC
'high'
that was also
message rate profile
used in
the flight simulator experiments described in the next chapter.
The
The high rate profile is
on the
(for reference)
included
timeline of figure B. 1, but is
Appendix C.
not shown
this profile is
dialogue for
in
included here strictly
for discussion and comparison with the 'low'
rate profile.
The Cruise Descent Segment
ATC message rate relative to
Figure 5.6 represents a low
Figure 5.6a shows
some of the data presented in
Chapter 4.
that during the long descent
this profile averages only one
message in
an 80 second
This corresponds
interval.
channel utilization (see figure 5.6b)
ment.
Examination of Panels 1-9
to a
of 8% during the seg-
of the timeline indicates
that messages are likely to cause few interruptions.
can be
From the timeline it
intercrew communications
cent.
Table 5.3
occurring during
(see columns 3 and 4)
communications are likely to
table
seen that there are also few
indicates that
interarrival time
tasks is unchanged if communications
-
134 -
des-
verifies that ATC
cause few interruptions.
mean
the
the cruise
The
between
events are included as
C
8
10
00
0
co
4-
0
FI NA L -I
DESCENT
Ti'me into Miss ion(sec.)
a) Smoothed Communications Message Rate
30
C
0'
4)
20
RMI NAL
TORING
DESCENT .|
4FINAL-4
10
5
100
1000
500
b) Communications Channel
Figure 5.6:
1500
Time into Mission(sec.)
Utilization
A Low Message Rate Communications Profile
-
135
-
DESCENT
100
-
TERMINAL
VECTORING
*1FINAL -
1500
Time into Mission(sec.)
a) Smoothed Communications Message Rate
500
1000
0
E
co
co
TERMINAL
DESCENT
100
.--+VE CTORING
500
-J|F1NAL
1000
1500
Time into Mission(sec.)
b) Communications Channel Utilization
Figure 5.7:
A High Message Rate Communications Profile
-
136 -
tasks that the
crew must monitor.
shows that the presence of ATC and intercrew
does not alter
figure 5,3
Similarly,
the degree of task
co munications
simultaneity during this
segment
The Terminal Vectoring Segment
It is e-vident
messages is much
during the cruise
from figure 5.6 that the
intensity of ATC
greater during the vectoring
descent.
segment than
The average message
rate is 7
messages in an 80 second period as compared to 1 per 80 secChannel utilization (see fig-
onds for the descent segment.
has increased from 8% for the descent to approxi-
ure 5.6b)
mately 30% for the vectoring
however,
segment.
that the message rate
is
It
still
should be noted,
quite 'low'.
figure 5.7a shows a
communications profile of
The
message rate
of nearly 14 messages per 80 second interval, twice that for
the
'low'
that these figures include
Recall
rate profile.
only ATC messages (messages to/from ATC by/for any aircraft)
Nevertheless,
and no intercrew communications.
the loading
effects of voice communications are expected to be much more
significant
during the
vectoring segment
than during
the
cruise descent.
Panels
9-14
of
the tiweline
indicate
that
required
intercrew communications associated with checklist execution
are numerous but of small average duration.
-
137
-
These intercrew
- On,' or 'Shoulder Harness -
tasks such as 'Seat Belt Sign
As intercrew communications become more numerous
Fasten.'.
they,
associated cockpit
trigger an
tend to
communications also
will tend to interrupt other tasks,
too,
disrupt men-
tal planning tasks and thus implicitly load the crew through
the same mechanism as ATC messages
5.3 shows
Table
effects
the
communications
of voice
on the nission mean task interarri-
(both ATC and intercrew)
val time.
In this table each
crete task
in addition
message is counted as a disother four
to the
task categories
Voice communication mes-
5.2.
that are tabulated in Table
sages are
(see Chapter 4.4).
separate category of
taken to be a
task because
the crew's attention is
diverted from other tasks,
to monitor a message.
Also,
in order
messages to which a crewmember
must respond require physical effort to grasp the microphone
and to speak.
sages as
from an
Table 5.3 shows that the
inclusion of mes-
Mission task
interarrival time
tasks reduces the
average of 17 seconds
between tasks..
tasks and
If only
pilot tmust participate
between tasks to
strong effect on
the presence
are
15 seconds between tasks.
of voice communications lias a
the magnitude of the
during the vectoring segment.
-
which the
standard flight procedure)
(per
included, the interarrival time is
In either case,
nessages in
13 seconds
133 -
interruption process
Another
effect caused
crew communications
5. 4b.
is
by the presence of
seen by
The figure indicates
ATC and inter-
comparing figures
that,
if
5.4a and
communications
tasks
are counted along with the other four task categories of the
timeline,
then the degree of
From figure
affected.
sion of
task simultaneity is
5.4b it
is
messages significantly
strongly
apparent that the inclu-
increases the
frequency of
occurrence of three and four simultaneous tasks.
,
may perceive himself
thus,
more effort
due to the
The pilot
to be incrementally exerting
influence of voice
messages during
this segment.
Examination of figure
few ATC
messages that
explicitly
B.2 (Panels 2-5)
are directed to
load the crew.
shows
that the
Clipper 54
tend to
'This is because the critical path
lies along the communication event path for most of the vectoring segment.
Most
of these messages invoke
Explicit loading due to
descent maneuvers to be executed.
ATC messages is
very evident,
turning or
then,
during
the vectoring
segment.
Following the above discussion,
how the high
it
is
message rate profile shown in
possible to infer
figure 5.7 will
affect crew loading during the vectoring segment.
rate
will cause
arrival of
ATC
the interruption
process
messages and the necessity
-
139
-
The high
induced by
the
to monitor them,
ATC Voice Communications mechanism will
the crew due to the
with the low rate profile
be greater than
in
(and in
5.6
figure
explicit loading due
messages
because
implicit mental loading of
The
to be greatly intensified.
to
The
the
aircraft
subject
are
scenarios described by fig-
dialogue which figure 5.7 represents
by introducing additional messages
is produced
of
not be affected
to ATC messages would
directed
depicted
The magnitude
the timeline).
unchanged in the communications
ure 5.6 and 5.7.
that is
directed to
traffic in the terminal area, other than Clipper 54.
The Final Approach Segment
average message rate at 10
Figure 5.6 indicates that the
messages per
during the
approach segment than
the final
channel utilization for
ment.
The
nearly
identical
at
30%.
higher during
is -slightly
80 second interval
This
vectoring seg-
these two
suggests
segments is
that
messages
encountered on the Tower control frequency are slightly more
frequent,
but of shorter mean duration.
According to Panels 14-16 of the timeline,
messages occurring are directed to
aircraft.
few of the ATC
Clipper 54,
the subject
This is substantiated by Panels 6 and 7 of figure
B.2 which shows that, unlike the vectoring segment, the critical event path
event path.
fore,
does not ccincide
with
the communications
Explicit loading due to ATC messages is,
likely to be minimal.
-
140
-
ther-
Implicit loading effects
sages are expected to be
tions are counted
intercrew. mes-
significant during this segment of
Again Iable 5.3 indicates
the arrival.
time is
due to ArC and
as tasks,
that when communica-
the mission
task interarrival
reduced from 16 to 12 seconds between tasks.
high rate
segment,
profile of
into the
were introduced
figure 5.7
If the
messages would be greatly
implicit loading due to
This increase woulld tend to increase the nomi-
increased.
for the segment as the
nal workload level
interruptions due
to messages increased.
magnitude of the
Note
that these
interruptions occur when there are two simultaneous tracking
tasks being performed
task slack
and when the timeline
indic'ates that
(and thus tasks are tightly
values are very small
constrained to be executed promptly).
Again,
the simultaneity histogram of figure 5.5 provides
additional information with regards to voice communications.
Comparison of figures 5,5a and 5.5b shows that the inclusion
of voice
messages as
shape of the
simultaneity histogram.
communications task
and five
of occurrence
tasks is reduced.
markedly changes
The presence
category causes the frequency
simultaneous tasks to be
The frequency
of voice
discrete tasks
of two
of the
of three
significantly increased.
and four
simultaneous
As in the vectoring segment, the presence
communications on
final approach
incrementally increase perceived workload.
this increment is
the
should tend
to
The magnitude of
investigated experimentally by introducing
-
141
-
the higher message rate communications profile of figure 5.7
(see Chapter 6).
CCNCLUDING DISCUSSION
5.4
The previous sections have described
nature,
of cockpit task loads during a simulated
a priori,
was then analyzed
This task load
Boston arrival.
context of the
in some detail the
workload elements discussed in
in the
chapter two.
It was found that the -task load and workload characteristics
of the three arrival segments
vectoring,
final approach)
(cruise descent,
terminal area
are quite different,
necessitat-
ing the separate consideration of these segments in workload
analysis.
This
section summarizes the implication
of the
preceding analysis and serves to tie the chapter together.
It
should be
noted that the phrase
'workload
been used in the relative sense in this chapter.
ario portrayed in this analysis
IFR,
is
transport category arrival.
workload is
has
The scen-
modelled after a normal,
In
probably never excessive.
some baseline,
level'
this nominal case the
However,
compared to
the nominal workload level can be inferred to
be relatively higher or lower
over different time intervals
of an arrival.
The Cruise Descent
-
142
-
The cruise descent was shown to be a rather long segment
where task
loading was minimal as were interruptions to this
this segment but
slack as to
these are afforded sufficient
allow ample time in
which to perform these tasks satisfactoto be relatively low,
Workload levels were inferred
rily.
occur during
tasks do
Several major planning
task load.
with the exception that high ATC messages rates may signifiworkload through the impli-
cantly increase perceived pilot
cit loading mechanism discussed in Chapter 4.
The 'erminal Vectoring Segment
shown to be half the
The terminal vectoring segment was
duration but
Interruptions caused
number of tasks performea per second.
by the
occurrence of multiple
shown to
decrease the mean
seconds,
for the descent,
planning tasks
and simultaneous
task interarrival
to
terms of
task load in
with roughly twice the
17 seconds.
tasks were
time
from 31
Navigation and
comprising approximately
were significant,
30% of the total number of tasks during the segment.
Voice
communications significantly
contribute to
crew
The task precedence map
loading in
the vectoring segment.
of figure
D.2 indicates that ATC
communications explicitly
trigger numerous tasks to be executed.
Intercrew commnunica-
tions associated with checklist execution also introduces an
-
143
-
loading effect by reducing
additional implicit
mean task
13 seconds from
interarrival time to
the mission
17 seconds
(see Table 5.3).
The nominal workload level would appear to be relatively
higher than that for the cruise descent segment based on the
analysis of the chapter and
the theory discussed in chapter
two.
the segment are still generous,
The slack values for
values during cruise descent.
though markedly less than the
It
is
suggested
values may result
that the presence of
in
smaller average slack
high workloads should
a cockpit emer-
gency or difficulty occur.
The Final Approach Segment
The final approach
vectoring segment
was shown to be very
in terms
of busyness
similar to the
and interruptions.
The notable difference between the two segments is that most
tasks which occur on 'final'
are associated with events that
lie along the critical path of figure B.2.
have little
time flexibility and must be executed promptly.
This segment
contains the
associated with flying the
simultaneous tracking
Instrument Landing System.
analysis presented provides no information
culty of
These tasks thus
this task
but
does suggest
-
144
-
tasks
The
as to the diffi-
that the
presence of
high ATC message rates, intercrew communications,
and other
required operating tasks can induce relatively high workload
or introduced
approach,
instantaneous task
It
levels.
from
'final' is
by external
workload
data
presented
cause
above satisfactory
level during 'final'
that during the cruise descent.
the
during
the final
sources could
demand rates to rise
appears that the
is higher than
little slack
extraneous tasks deferred until
final approach,
clear
are afforded
Because tasks
levels.
whether
It is not
workload
during
relatively higher or lower than the level for the
vectoring segment.
-
l4s -
Chapter VI
FLIGHT SIMULATOR EVALUATICN OF A SUBJECTIVE RATING SCALE
This
chapter
a series
describes
flight
of
simulator
a preliminary test of the
experiments that were designed as
11IT, subjective pilot workload rating scale.
EXPERIMENTAL OBJECTIVES
6.1
described in this chapter
The experiments
nary in nature.
were prelimi-
workload scale is
Because the MIT
has undergone only elementary design iterarions,
new and
the struc-
ture of the scale is expected to evolve, to some extent,
as
The properties of rat-
it is used by engineers and pilots.
ing scales are strongly influenced by errors of construction
and implementation.
step
in
Experimental use
developing the
accepted form.
scale
rating
is
thus an important
into a
useful
and
Such 'development' is the primary objective
of the experiments described herein.
Next,
this thesis has described some relatively new con-
cepts in the theory of
described in this
these concepts.
pilot workload.
The simulator test
chapter were designed to
the following issues were
In particular,
investigated experimentally:
-
146
explore some of
-
1.
Sensitivity of the scale versus the
performance
of
sensitivity
mea-
sures.
2.
among
of ratings
Consistency
across subjects for
and
the same nomi-
nal task load.
3.
The
relationship between
busyness
and perceived workload.
4.
to
the scale
of
Sensitivity
implicit loading of
the
ATC communica-
tions.
6.2
FACILITY
FLIGHT SIMULATICN
Flight missions were simulated using
707 simulator.
Figures
a fixed base Boeing
6.1 and 6.2 show
the exterior and
interior of the aircraft cab.
Although the cab mockup is
substantially accurate replica
of the 707,
station is
configured
no flight engineer's
without flight controls and
from figures 6.1 and 6.2 that
there is
is
Again,
station.
it
should
the cockpit is
thus no physical cockpit motion.
no provision for out-the-window
are some
the co-pilot's
particular,
In
noticeable features missing.
there
a
there is
be noted
fixed base and
Moreover,
there
visual simulation so all
simulated flights represented instrument meteorological conditions (IMC).
-
147
-
Figure 6.1:
Figure 6.2:
Cab Mock-Up
Cab Mock-Up Interior
-
148
-
Flight Instruments
The flight instruments
computer and displayed
were generated in real
on cathode ray tubes
center of the captain's and first
time by a
located in the
officer's panel.
The com-
puter generated flight instruments moved in a realistic manner and were displayed essentially flicker free.
seen from figure 6.2, airspeed,
FMI7,
altimeter,
attitude director indicator,
indicator were
the Collins
generated.
and
vertical speed,
aorizontal
The ADI/HSI was
FD-109 flight director.
experiments only raw data (Flight
As can be
situation
modelled after
However,
during the
Director Mode - Off)
was
displayed and so no command bars were generated.
The Simulation
The
facility's simulation
capability
Adage AGT-30 minicomputer which was
cent to the cockpit area.
functions in
1,
was
located in
based on
a room adja-
The Adage computer served three
the simulation process:
It simulated
the equations of motion
of the
Boeing 707.
2.
It
sampled
and stored
control inputs
from the
flight performance data
cockpit
from the
experiments.
3,
It
generated
an
the flight,
as the simulation ran.
149 -
instrument displays
per second and a new
Cockpit controls were sampled 15 times
aircraft state
was computed 30
times per
second.
Flight
instrument displays were updated 5 times per second. -
loop
closed
computer form
Adage
and the
the cockpit
Together
used
simulation capability
flight
in
the
these
High computation speeds allowed for good res-
experiments.
ponse to pilot control inputs.
Crew Statior.s
For this
Therefore,
test series the
only the
captain's
panel switches,
ries were
operating
station
can be seen
6.2 it
From figure
detail.
captain was the
placards,
It
guite complete.
test subject.
was outfitted
to
that the overhead
and functional accesso-
was felt that
attention to
such detail was important in simulating the cockpit environSeemingly unimportant hand move-
ment for workload studies.
that are important elements
ments all comprise distractions
the 'loading'
of
process in
detail was attempted in
6.3
EXPERIMENTAL
Two basic flight
evaluation tests.
6.4)
cruise
involved
the
cockpit.
Attention
preparing this experimental
to
study.
SCENARIOS
profiles were used
Each flight profile
an instrument arrival
configuration,
an
through
Boston's Logan airport.
Scenario
to form five unique tests.
-
150
-
in
the rating scale
(see figures 6.3 and
from a
high altitude
instrument approach
to
'parameters were adjusted
Albany, NY
Gardner
VOR
118 deg.
magnetic course
' MANJO
Intersection
Boston
Lo an
S11
160
035
MILT
LOM
Plan View (Courses shown are magnetic)
a)
36000 ft
22000
6000
100 ft
b)
Altitude Profile
Figure 6.3:
ILS 4R Arrival Profile
-
151
-
Albany, NY
VAR
9-
Gardner
VOR
MANJO
In ters.
118 degrees
magnetic course
15.0
DME-Goston
090
Boston
Logan
a) - Plan View (Courses shown are magnetic)
Albany, NY
VR 36000 ft
4000
Boston
.Logan
b)
Altitude Profile
Figure 6.4:
VOR-DME 15R Arrival Profile
-
152 -
The basic
profile shown
until the
federal airways
reached the Boston terminal area at
tored to the
6.3 consisted
New York at 36000 feet.
arrival frowF Albany,
routed along
in figure
This first
to Logan's runway 4R.
subject aircraft
The flight was
simulated aircraft
which point it
final approach course of the
Logan airport.
of an
was vec-
active runway at
profile included an ILS approach
mission was terminated when the
'The
published Category
reached the
II missed
approach point.
(that of figure 6.4) was nearly
The second flight profile
identical to the
first.
from Albany, N.Y.,
It consisted of
but concluded, instead, with vectors to a
VOR-DME approach to runway 15R.
minated upon
the same arrival
the arrival of
Again,
the flight was ter-
the simulated aircraft
at the
published missed approach point.
Workload Parameters
Each of the
above flight profiles was altered
in one of
three ways to increment the nominal workload level:
1.
Increment
the
A'C
message
rate
(i.e.,
the rate at which radio mes-
sages
occur
Figures 6.5
on
the
frequency).
and 6.6 show
the high
and low rate message profiles which
were used.
-
1 r3
-
2.
Execute the flight
profile of figor
ure 6.4 rather than that of 6.3,
vice-versa.
3.
of
unexpected change
Introduce an
the instrument approach,
just prior
the approach
to initiating
proce-
dure.
Altering the workload level in this manner had the advantage that no
artificial tasks (such as
were introduced,
cancelling task)
a secondary,
que eiliminated
it
the intrusion effects of
This techni-
secondary loading
The disadvantage of this technique, however,
methodologies.
was that
control was
while some
maintained over the workload increment.
still
light
was
create workload
difficult to
levels that
This may have caused difficul-
were incrementally 'higher'.
ties in exercising the sensitivity of the measure.
present objective,
the
though,
value of
For the
using realistic
loading mechanisms was thought to be important in evaluating
a workload measure for the air transport environment.
Five unique test scenarios were
thus defined.
The sce-
narios are summarized in Table 6.1 and a test matrix that is
helpful for identifying
6.7.
these scenarios is shown
These five experiments
allowed the hypotheses previ-
ously outlined to be examined more closely.
-
154
in figure
-
(U
CO
0
F INAL
DESCENT
-Ai
Time into Mission(sec.)
a) Smoothed Communications Message Rate
0,
30
(A
E
C
Cn
a.0.
0
20
U
Cn
to
DES CENT
MI NAL
0 R ING -
-|V
.
F INAL-4
U,
10
5
1000
500
100
b) Communications Channel
Figure 6.5:
.
1500
Time into Mission (sec.)
Utilization
Low Message Rate Communications Profile
-
155
-
DESCENT
100
TERMINAL
--imVECTORING1
500
FINAL
4
1500
Time into Mission(sec.)
a) Smoothed Communications Message Rate
1000
C)
8,
QU
eo.
en
&9
DESCENT
100
..
TERMINAL
VECTORING
F INAL M
1000
500
1500
Time into Mission(sec.)
b) Communications Channel Utilization
Figure 6.6:
High Message Rate Communications Profile
-
156
-
EXPERIMENTAL PROCEDURE
6.4
Test subjects
their participation in
1.
Preflight Briefing
2.
Warmup Flights
3.
Data Flights
4.
Post Flight Critique
MIT scale as
part of
the rating scale evaluation.
These experimental
scale's use
step process as
completed a four
flights were aimed at
a measure of the pilot's
for other
crew stations
All simulated flights situated
evaluating the
workload only.
was not
tested here.
the subject as pilot-in-commanual flight.
No
or flight director information was provided.
It
mand and pilot-at-controls in
The
auto pilot
was,
ever, necessary to simulate the captain's duties,
how-
his cock-
pit station, and the captain's interaction with his crew.
As has been mentioned,- the
cated to detail,
was
no flight
duties with
formed by
captain's station was repli-
the copilot's
engineer's
which a
was less detailed and there
station.
captain normally
the copilot.
Primary
Also,
there are,
engineer
interacts were
examples of
were preparation of landing/approach
checklists.
The flight
per-
these duties
data and the reading of
in actual line operations,
intercrew communications that take place between captain and
-
157 -
..
........
Table 6.1:
Scenario Summary
Scenario
No.
Description
Albany arrival to an ILS runway 4R.
ATC message profile.
Low
Albany arrival to an ILS runway 4R.
ATC message profile.
High
Albany arrival to VOR-DME 15R.
ATC message profile.
Low
Albany arrival with a planned approach
to an ILS runway 4R. Approach switched
to VOR-DME 15R, just prior to MANJO
intersection. Low ATC message profile.
Albany arrival with a planned approach
to a VOR-DME 15R. Approach switched
to ILS runway 4R, just prior to MANJO
intersection. High ATC message profile.
-
158
-?
ATC Message Profile
.. '
lit
Yes
Change
in
App roach?
ATC Message Profil'e
Low
High
Albany Arrival
ILS 4R
Scenario 2
Scenario 1
Scenario 5
-
Scenario 3
Albany Arrival
VOR-DME 15R
Figure 6.7:
1
Scenario 4
Scenario Test Matrix
- I ;C) -
The experiments
were not simulated.
These communications
flight engineer.
thus replicated
a two-pilot
crew such
as
currently exists on the Boeing 737 and DC-9 aircraft.
the pilot subjects acted as both pilot-in-command
Again,
and as pilot-at-controls for the entire duration of the test
scenarios.
for by the
The copilot
was told to perform
those commanded by
flight operations manual and
the captain-subject.
duties called
The copilot could not, however,
acti-
vate the control wheel or rudder.
6,4.1
Preflight Briefing
Prior to flying the simulator,
to fill
out an
Basically this
subject pilots were asked
experience questionaire
(see Appendix
C).
questionnaire identified the -subject's cur-
rency and general experience in IFP and air transport operations.
Pilots were next briefed on the purpose of the experiment
and given
assessment
briefing,
a detailed
scale.
briefing on the
Upon
completion
use of
of
the workload
the rating
scale
subject pilots were then briefed and familiarized
with the MIT simulation facility.
Each pilot was supplied
with a set of instrument enroute and approach charts for the
New England
area.
Pilots
unfamiliar with
given the opportunity to review these charts.
-
160 -
the area
were
subjects were
Test
procedures.
crew
These procedures were drawn from the operations
that operates the Boeing 707-320
airline
manual of a U.S.
aircraft.
set of
a standard
briefed on
The procedures were thus representative of stanbriefed on
Each pilot was also
dard operating practices.
and as
as pilot-in-command
his responsibilites
and duties
pilot-at-controls.
Subjects were briefed, in addition,
whose role
the copilot,
the duties and responsibilities of
on
was played by an MIT researcher experienced as an IFR pilot.
should be noted that the majority of the participating
It
subject
this evaluation were Boston
pilots in
subjects were,
for
the most part,
Boston area and with the
Logan airport.
and so
very
based.
The
familiar with the
standard instrument approaches for
subject pilots were 707 pilots
None of the
the characteristics of
the simulated
aircraft were
unfamiliar, initially, to all.
The preflight briefing required an average of one hour to
complete (excluding the questionnaire).
6.4.2
Warwup Flights
Test subjects flew a series of 'warmup'
tom the pilots to the flight procedures
characteristics of the simulator.
up
scenarios in
flown.
the
and to the operating
Table 6.2 lists
the warm
they were
generally
which
order in
flights to accus-
warmup flying period was reduced
In some cases the
-
161
-
I-
particularly adept at flying the
as some of the pilots were
simulator.
In
other cases the
subject pilot
until the
warmup phase
seemed at
ease and
was extended
performing to
nominal IFR standards of acceptability.
TABLE 6.2
Warmup Flight Scenarios
Arrival from Gardner VOR at 14000 feet with vectors
i)
to an ILS approach to runway 4R.
ii)
iii)
Approaches began
Two practice approaches on the ILS 4R.
at 2000 feet, 2 miles west of the final approach course, on an
intercept heading of 70 degrees.
Full VOR-DME approach to runway 27. Approach
begins south of the Boston VOR at 2000 feet.
Note:
6.4.3
A low ATC message rate was used during the warmup.
Data Flights
Subject pilots flew three arrivals
collected.
during which data was
In some instances fewer than three arrivals were
flown due to equipment malfunction.
data was sampled in a wanner
comparisons to
In all cases, however,
that allowed a variety of data
Workload data
be made.
that was
sampled
could be compared across measure types, across pilot groups,
between flight segments,
and across test scenarios.
Flimht Data: Performance
-
162
-
The pilot's mean
absolute deviation
precision approaches)
and frcm localizer
segment of each arrival.
(on all approaches)
It
has
been shown that,
in
some
performance data is sensitive to variations in total
Performance data could thus be used in a compari-
workload.
tive
path (on
computer during the final approach
was sampled by the Adage
cases,
from glide
manner
with
the
rating
scale
results
for
final
approach.
(physical control activity)
The displacement of controls
was sampled and recorded on
data flights.
Controls sampled were aileron, elevator,
pitch trim.
activity
load.
an analog chart recorder during
it was hypothesized
that the
and
flight control
(ir.manual flight) might serve as an index of work-
This data, then serves as another measure for compar-
itive use.
Flight Data: Subjective
Ratings
Each data flight consisted,
These were the cruise descent,
segments.
toring,
and final approach segments.
val, these segments had durations of 14,
subject was
an arrival,
asked to
average segment
For the ILS 4R arri8,
and 4 minutes,
4 and 4 minutes.
segments had durations of 14,
segment of
terminal area vec-
for the VCR-D E 15R arrival,
Similarly,
respectively.
of three mission
logically,
use the
rating scale
minimize discontinuity effects.
-
163 -
and the
to assess
Freeze 'points were
workload.
After each
simulation was frozen
the
these
his
selected to
In addition
pilot's workload.
lot/observer also rated the subject
was an atterpt to investigate
vers to rate
the copi-
pilot's self-assessment,
to the
workload.
It
This
the ability of trained obsera cross-check on
also provided
the ability of a crewmember to evaluate his own workload.
Flight Data: Video
A low light-level
pilot's hands
camera's
over the captain's
Control
field-of-view of the camera
brake,
right shoulder so
and flight instrument
field-of-view.
The camera
the cockpit.
record audio and video activity in
was focused
all flights to
camera was used during
(such
group were
movements
that the
within the
outside
as altitude alert,
the
speed
and throttle) could also be noted.
Several subjects were
asked to view the
video replay of
after completion of this flight.
one of their test flights,
Again the video was frozen at the end of a segment of interest,
at
was asked to
which time the subject
segment workload.
reassess the
This data was expected to yield informa-
tion on the ability of
pilots to self-assess their workload
and to experiment with post-test, pilot calibration*.
*Calibration would involve telling a
flight represented a two not a four.
-
164
-
subject that
this
DISCUSSION OF RESULTS
6.5
In total 8 pilots participated in
ation tests.
Of
the rating scale evalu-
these four were airline
pilots with over
4000 hours of air transport experience and four were general
aviation pilots,
aviation,
with less than 1003
two
This sample size proved
IFE experience.
adequate for
identifying some properties
some problems associated with its
provide a
hypotheses.
data base
major rating scale
to be
scale and
As these
the sample was not meant
with which
The discussion
of the
implementation.
tests were preliminary in nature,
to
hours of general
that
to statistically
test
follows highlights
the
properties and problems that
were noted
in testing.
The Learning Curve
The
prixary difficulty
experiments was a marked
by the subject pilots.
a particular flight
jects became
the
encountered
in these
simulator
learning curve effect demonstrated
That is, pilot workload ratings for
phase consistently dropped as
intimately comfortable with the
flight procedures.
the sub-
simulator and
occurred despite
This
concerted
efforts to circumvent learning curve effects by establishing
a rather long briefing/warmup flight
section 6.3.
Pilots
spent approximately six hours
MIT facility during their
mately three
period as described in
participation.
hour of this three was spent
-
Of this approxi-
flying the simulator
hours was spent
in
165
warmup flights.
-
at the
and one
The result
of the
pronounced learning curve
was that consistency,
within a subject,
much of the
rating data.
later runs,
afte&
It
is
evident
effects on the
was not evident in
Consistency was
the learning curve had
from figure 6.8a that the
evident only in
'flattened out'.
same scenario con-
sistently produced lower ratings in later runs,
subject pilot.
Morever,
test data
for a given
figure 6.8b indicates that, in sev-
eral cases, these lower workload ratings were accompanied by
better performance.
Together,
these data confirm the pres-
ence of strong learing curve effects.
These learning curve effects
can be drawn from the rating
tant
to
note.
It
strongly influence
scale.
is
limit the conclusions which
scale data and thus are impor-
believed
that such
other experiments
The following conclusions can
effects
utilizing the
will
rating
be drawn with regard
to learning curve effects and the use of rating scales:
1.
If consistency
desired,
across or within
the
experiment will
repeated until such time
bilize.
have
to
be
as the ratings sta-
Performance stabilization and verbal
questionnaires are
inadequate indicators
learning curve plateaus.
well
subjects is
show excellent
6.8)
-
166
Subjects may very
performance but
lower and lower workload
-
of
assess
ratings (see figure
FA
2
[]
FA - General Aviation
Pilot.
Final Approach
02
TV - Airline Pilot.
Terminal Vectoring
Suoerscript indicates
subject- nr.
FA
2
Run No.
a) Workload vs.
Run No.
#2
#1
FA
Subject iNo.
-
5
FA - Subject t
#
#1
FA
-
Subject 3
d#2
100
b) Workload vs.
Figure 6.8:
200
-
Indicates Run No.
)-
Indicates Ai rline
Subject
[]-
Indicates General
Aviation Subject
400
300
RMS Deviation from
PLS 4R
Performance
Learning Curve Effects
-
167
-
If
2.
unfamiliarity
or surprise is an
the measurement experiment,
is
desireable in the
issue in
but consistency
rating lata,
ject pilots must be
then sub-
chosen with very similar
experience and skill levels.
It
that pilots
emergency work-
in this kind of
measurement
-load
environment,
extensive training
Such
training
also appears
should
in the use of
woUld
have
the scale.
necessarily
calibrate
the subjects to the rating scale.
Egocentricity
Second in
importance to learning
rating scale data is
tricity.
Flight
curve problems
in the
the apparent probLem of subject egocen-
officers and
pilots,
sect, have very protected egos.
as
a sociological
As a group they seem parti-
cularly proud of their skills and their flight positions and
thus seem
to be cautious in
truthfully assess
the extent to which
their workload.
experiments encountered one
In particular,
subject who would not
test situation higher than the
they will
these
rate any
lowest conceivable rating (a
one or a two).
Figure
6.9 shows
'outlier' for
that
subject number
the final approach
3
is indeed
scenario that
an
involved a
VOR-DME approach to runway 15, with a low message rate.
For
this run the subject assessed his workload as being very low
-
168 -
4
0
30
O
3
Egocentric
.C_
Outlier
(D
-o
02
.
2
0
400.
'800
1600
1200
2000
2200
Mean Absolute Devjat ion from
Final Approach (f et)
Figure 6.9:
Effects of Egocentricity
and, yet,
his performance was well outside the tolerances of
instrument flight rules.
the subject's multi
Also,
for this particular segment,
attribute rankings state that
the sub-
ject perceived only occasional moments of free time,
ate levels --of-information
processing intensity,
levels of emotional stress and confusion.
seem-incon-sistent wi-th a run where
moderand mild
These perceptions
the sub ject was one half
a mile off course and his rated workload as low as a 2.
In
these experiments,
egocentricity appeared
problem with only one subject and
therefore not as
Despite this,
it
its
a
affect on the data is
significant as the learning
is
to be
curve effect.
thought that egocentricity
may pose a
difficulty in future use of the pilot workload rating scale.
These tests
seemed to
indicate that
further ego
problems
could be eliminated by including the following steps as part
of the preflight briefing:
1.
Ensure that
pilots understand that
assessing the
workload reguired by a
flight
procedures
flight
environwent.
that are
they are
in a
It
given
is
being evaluated,
the
set of
cockpit
and
procedures
not the
subject
pilot.
2.
The subject
of how
ment,
pilots should not be
'most'
pilots rate a
unless the intent
-
170
-
is
made aware
particular segto calibrate the
subjects to the scale.
also not
if
The subjects should
be unintentionally made to
being compared
they are
to
feel as
a pilot
of
higher or lower skill/experience/prestige.
These steps seemed to have
rinimized further data integrity
flaws due to subject egocentricity.
Workload Perception:
One of the
Self-Assessment vs.
objectives of this test
question marks associated with the
scales,
concerns the
his own workload.
general,
ver*,
assessment
was
indeed,
use of subjective rating
pilot to assess
Figure 6.10 indicates that there was, in
displays
The figure
of this
series and,
ability of a subject
excellent agreement
with most
Cbserver Assessment
between
subject and
that the
lower than the subject's
observer's
assessment,
Alt-
does not prove that subjects can
hough figure 6.10 certainly
assess their own workload,
scatter,
surprisingly little
scatter indicating
obser-
it
does lend some supportive evi-
dence to this hypothesis.
One subject
pilot was
scale so that he was,
nominal
assessment as
results of this run.
was good
familiar enough
at cued intervals,
he
flew.
with the
rating
able to call out a
Figure 6.11
shows
the
It should be noted that, again, there
agreement between subject and
observer**.
Also,
*Observer and subject assessments were made independently.
**Note that the observer's assessment was made upon viewing
-
171 -
0
0
0
0
411
Subject High/
Observer Low
1
Subject Low/
Observer High
2
Observer's Rating
a) Airline Subjects
ISubject
High/
4Observer Low
Subject Low/
Observer High
Observer's Rating
b) General Aviation Subjects
Figure 6.10:
Subject vs. Observer Ratings
the video replay.
-
172 -
of special interest, the jaggedness in the early portions of
the rating profile
( see figure 6. 11)
well to the
correspond
timeline of figure B.1 of
nominal task load depicted in the
the appendix.
reassess their workload using the
of their flights and then
scale.
case did these reassessments
In no
different
ratings and
remained unchanged.
in most
Finally,
whether they
debriefing,
a video replay
were asked to view
Several test subjects
cases
yield markedly
the original
rating
each subject was asked,
felt
it
during
to judge or
was possible
perceive their own workload.
The
was a unanimous affirmative,
which was generally expressed
with a strong
if
desire to try the
response to this inquiry
self assessment
to its value.
there was some doubt as
procedure,
This was perhaps
the strongest evidence in favor of the subjective assessment
It
technique.
of workload
indicates that pilots do seem (over the range
to feel
levels tested)
that they are
able to
rate their cwn workload.
Workload,
Performance,
Chapter 2 summarized
tional
Simpson and Sheridan's
relationships between
skill level.
is
and Skill Level
workload,
is
performance,
funcand
Simpson and Sheridan would contend that there
no simple relationship between
workload
(1978)
not
equivalent to
-
173
these parameters and that
performance.
-
Figure
6.8b
Cruise
Descent
Terminal
Vectoring
Final
Approach
I
I
I
A
1
a
I'
j
I
'
I
a
I
-J
Observer's Assessment
Subject's Assessment
Minutes into mission
Figure 6.11:
.
Continuous Workload Assessment for -an ILS 4R Arrival
~,b.
shows
examples
several
performance,
all
perceived
of
consistent with
of which are
versus
workload
the theory
outlined in chapter 2.
The general aviation pilot
points)
assessed
identical ratings for
2.
two runs
with the
and yet his performance improved by
same nominal task load,
a factor of
(indicated by the rectangular,
The two examples using
(the circular data points)
indicate
airline pilot data
both a workload reduc-
tion and a performance improvement in the later runs.
as the
pilots'
skill
increased their
performance improved
and their workload was rated as lower.
general aviation pilot,
as his skill
Thus
For the case of the
level in performing the
final approach increased, he was able to expend a comparable
amount of effort and yet
6.8b that the
Notice in
perform better.
general aviation pilot performed
airline pilot
as well as the
(on
figure
(on his 2nd
his earlier
run)
nearly
run),
but the general aviation subject rated his workload to
too, is
con-
be much higher than the airline subject.
This,
sistent with theory which states that,
the skill level of
one pilot is
higher than that
if
of another,
the more skilled
pilot will exert less effort to perform the same tasks*.
Workload and Busyness
*In
this case the general aviation subject's experience,
is a fraction- of that of the airline
measured in hours,
Hours experience was taken here to be a proxy
subject's.
although flight hours is not always a good
for skill,
metric of relative skill.
-
175
-
It
is
of busy time'
interesting to note . that 'percentage
appears to
functional relationship
bear no simple
That is, the data indicates numerous exam-
ceived workload.
both the task timeline of
ples of low workload levels where
figure B.1 and
indicate high
-- cases-were
the multi attribute ran-kings
(see table 6.3)
periods where
workload during
low levels of busyness.
these
low level effort.
there are also examples
activity tiueline and the
in
instrument approach
a-f-airy-difficult
maneuver using almost reflexive,
relatively high
of the subject
The subjects
levels of busyness.
perform
Conversely,
to per-
of
the cockpit
multi attribute rankings indicate
These data provide further suppor-
tive evidence to the theory outlined in chapter 2 which suggests that busyness, alone,
is not a good workload measure,
even in an aircraft under manual control.
Rating Consistency within a Subject
Due to the learning curve
cussed previously,
and egocentricity effects dis-
the simulator tests
did not
yield any
reliable data on the repeatibility of ratings for equivalent
nominal task loads,
within a subject.
Rating Consistency across Sutlects
Ignoring the t'est subjects' earlier
curve effects,
there is
some encouraging evidence that con-
sistent ratings, across pilots,
shows
rating
data for
runs due to learning
the
-
are attainable.
high
176 -
ATC message
Figure 6.12
rate,
ILS
Table 6.3:
Workload Ratings vs. Multi attribute Busyness
Rankings
Busyness
Attribute
Workload
Rating
Always
Idle
..
Have Free Moments
often
Occas.
Rarely
Always Fully
Occupied
4
x
3
x
2
x
2
x
3
x
4
x
x
iii
x
2
x
3
4
x
iv
x
5
4
o
4
5
3
<
C3
x
x
x
x
x
S4
iL
x
approach to runway 4R.
data points
Observing
(the circular points),
ratings are fairly consistent.
eral concensus
only the airline subject
it
is
The figure indicates a gen-
that the cruise descent
terminal vectoring
stage is
apparent that these
a 2
or 3,
phase is a
1,
the
and that--the final
approach is a 2 or 3.
Eecall that these rankings are
results of the
quite consistent with the
analysis of chapter 5.
gested that the
cruise descent phase of
induce the lowest workload levels
ments.
The
analysis also
whether the
ment.
the arrival should
of the three arrival seg-
concluded that
nominal workload
lower than that
higher or
This analysis sug-
on final
it was
unclear
approach should
for the terminal
be
vectoring seg-
Note also that all segments seem to have a fairly low
nominal workload.
of chapter 2
This,
too,
is consistent with the theory
and the analysis presented in
reconfirms what is already common knowledge.
nal workloads
excessive,
on routine,
air
chapter 5.
That is, nomi-
transport arrivals
despite the complexity
It
are not
and busyness depicted in
figure B.1 of the appendix.
Observation of the general aviation subject ratings indicates that
In
subject.
in
part,
consistency is
part,
less obvious
this is
for this
due to the small sample size and
to a greater degree of data scatter.
in scatter
is very
likely due
-
class of
178
to 'the
-
This increase
wider variation
in
0
0
05
0-
0Cruise Descent
Figure 6.12:
Terminal
Vectoring
Rating Consistency for an ILS 4R arrival
Indicates General
Aviation Subject
Indicates Airline
Subject
Final
Approach
experience and skill that is
pilots,
than is
Cne trend
present among general aviation
present among Part 121,
that is
evident in
airline pilots.
figure
6.12 is
that the
general aviation subjects seemed to have, on average,
the workload
for this scenario to
transport pilots.
be higher than
This lends some
skill/performance/workload
rated
have the
further support for the
relationships
discussed
previ-
ously.
To summarize, the data does provide evidence showing that
consistent ratings are attainable.
however,
cularly
The small sample size,
limits the conclusiveness of this evidence,
parti-
with regard to the repeatability of a subject's rat-
ing.
Sensitivity of the Scale to ATC
tssg
Rate
These tests provided no reliable indication that the rating scale
was sensitive to an
increase in
the implicit loading effects of
learning curve
workload
ATC messages.
Once again,
effects occurred across scenarios
designed to investigate implicit loading,
due to
that were
causing this data
to be unuseable for this purpose.
Pilots did,
message
however,
scenarios.
react on
In
several occasions to the
one instance,
insisted on knowing the distance
a
subject
pilot
of his aircraft from other
simulated aircraft and several times- prompted the controller
-
180
-
for this information.
ject had a
As discussed in chapter 4,
mental image of the local
this sub-
traffic situation and
knew that there were aircraft not far behind him in
the ter-
minal area.
As
another example
of implicit
thought one of their ATC
loads,
pilots
several
clearances to be ambiguous.
From
the clearance
Clipper 54,
depart MANJO at 6000
feet on a heading of
170 degrees
it
was not obvious to these pilots whether they were .cleared
down to 6000 feet,
generally
discussed
Following this,
the
clearance
these captains
itself
and,
in
some
examined navigation charts for minimum altitudes
instances,
or asked ATC
ance.
or not.
to confirm their interpretation
None of these extra tasks
lytic description
of a timeline
instantaneous workload
of the clear-
is noted on the task anaand all contribute
of the crewmember.
to the
Moreover,
the
confusion aroused by this message was apparent to the observer,
upon replay of the video record-
and to the subjects,
ing.
Despite
these
noted
occurrences
effects due to ATC messages,
any reliable evidence as to
effects.
Moreover,
experiments was such
of
implicit
the rating data do not provide
the scale's sensitivity to such
the control
of the
flight simulator
that the multi' attribute
provide no evidence of these effects.
-
loading
181
-
ratings also
6.6
CONCLUDING DISCUSSICN
ro
conclude, it is
important to restate that these flight
simulator evaluation tests scale were preliminary
They
were
not
intended to
provide
demonstrated the sensitivity and
Indeed,
the acceptance
air transport
final
in nature.
evidence
that
universality of the scale.
of a workload rating
scale for the
environment will necessarily -regnire experi-
mentation in high fidelity simulators
and in real aircraft.
Such tests will require input from many pilots,
ers, and operator.
manufactur-
Such acceptance was not the objective of
these experiments.
The experiments
described in
this chapter
great deal of valuable information on
the properties
noted that
of the MIT
scale.
learning curve
did yield
a
the proper use and on
In
particular,
effects caused
it
ratings to
was
drop
steadily as pilots became more
familiar with the simulator.
Performance,
showed little
at the same time,
ment.
Pilots thus required
verbal
statements
or no improve-
longer test sessions,
and performance
levels
that
despite
suggested
their mastery of the flight simulator/crew procedures.
Despite
effects,
mined.
limitations on the data imposed by learning curve
some
theoretical concepts were
Of special interest
were results which consistently
showed that perceived workload is
mance,
nor is
successfully exa-
not equivalent to perfor-
the level of busyness equivalent to perceived
-
182 -
The data showed instances of low workload ratings
workload.
showed high workload
100% busyness and it also
at times of
ratings at times of low busyness.
Surprising rating
an equivalent
for
consistency,
load/scenario, was achieved across pilots.
task
There was, how-
a noteable tendency for the general aviation subjects
ever,
higher workload ratings,
to assess
for a
given scenario,
than their air transport counterparts.
Due to the presence
of pronounced learning curve effects,
the test data did not
yield any reliable information on the sensitivity of the MIT
scale to
implicit loading by
mentioned,
that
though,
'incidents' were observed
test
rate,
scenarios.
ATC messages.
a number
of
implicit
to occur during the
It
was
It
loading
high message
simply not clear
available data whether the scale
should be
from the
was sensitive to the pres-
ence of this loading mechanism.
It can be concluded that
ciated with the
use of any subjective
for pilot workload.
the £II scale
It
tion.
there are indeed problems assoassessment technique
these experiments indicate that
Yet,
can provide valuable and
consistent informa-
further research be undertaken
is suggested that
using high fidelity flight simulation technology with experienced,
lize
These experiments must uti-
line crews as subjects.
structured scenarios that include
traffic control.
ATC imposes
-
the presence of air
perh-aps the most influential
183 -
constraints and is the predominant source of stochastic disturbances to the nominal task load.
The results of these tests indicate that only the low end
of the rating scale is
arrivals.
there
is
exercised by routine,
air transport
To exercise the full range of the workload scale,
no
doubt
that a
introduces abnormalities
structured
and emergencies will.
With this 'apparatus' as an
possible to
obtain the
consistency,
sensitivity,
scenario set
experimental base,
more conclusive
which
be required.
it would be
evidence on
scale
and acceptability that the experi-
ments described in this chapter suggest exist.
Finally,
the overwhelming
participating in
response of the subject pilots
these evaluation tests,
they could perceive their own workload.
expressed an eager
desire to try the
This finding may be the most
Moreover,
subjects
subjective technique.
important result of the rating
scale evaluation series.
-
was that they felt
184
-
BIBLIOGRAPHY
BabcockG. L. and Stone,E.B., Pilot Error - Past, Present
and Future Paper Presented to the Aviation Psychology
October 18, 1977.
Association San Francisco, CA
Beatty, Jackson, Pupil Dilation as an Index of Workload
Proceedings of the Symposium on Man-System Interface
Advances in Workload Study. Air Line Pilots Association.
Washington, D.C. July 31 and August 1, 1978.
Billings,C.E., Lauber,J.K., Cooper,G.E. , Ruffel-Smith,H.P.,
Retrospective Studies of Cperating Problems in Air
Transport Proceedings of NASA Conference: Aircraft Safety
and Operating Problems Hampton, VA.
October 18-20, 1976.
NASA SP-416.
Chiles,W. Dean and Alluisi,Earl A., On the §ecification of
Operator or occupational Workload with PerformanceMeasurement ethods Human Factors, 1979. pp 515-528.
Donchin, Emanuel, Brain Electrical Activity as an Index of
mental Workload in Man-Machine Systems Proceedings of the
Advances in Workload
Sysposium on Man System Interface:
Study. Air Line Pilots Association. Washington, D.C.
July 31 and August 1, 1978.
Dunlay,William J., and HoronjeffRobert, Application of
Hiuman Factors Data to Estimating Air Traffic Control
Conflicts Transportation Research. Vol 8, pp 205-217.
Great Britain 1974.
Pergamon Press:
Dunlay,William J., Stochastic Properties of Total Air
Traffic Control Voice Communication Time Transportation
Pergamon Press: Great
Research, Vol 9, pp 275-278.
Britain, 1978.
Edwards,A.L., Techniques of Attitude Scale Construction New
Appleton Century Crofts, 1957.
York:
Ellington,John E., Airline Approach to CAT III 22nd
Symposiufr Proceedings, Society of Experimental Test
Pilots, 1978 Report.
Sept. 27-30, 1978.
-
185
-
Ephrath, Arye, Pilot erf ormance in Zero Visibility
Approaches PhD Thesis. Department of Aeronautics and
Astronautics,
Massachusetts Institute of Technology.
1975.
Gartner,W.B. and MurphyM.E., Pilot Workload and Fatigue: A
Critical Survey of Concepts and Assessment Techniques
NASA TN D-8365 Ames Research Center, Moffett Field, CA
94035.
November 1976.
Federal Aviation Regulations Part 25.1523 and Part 25
Appendix D.
Hart,Sandra G. , Subjective Time Estimation as #Z Index of
Workload Proceedings of the Symposium on Man System
Interface: Advances in Workload Stud! Air Line Pilots
July 31 and August 1,
Association. Washington, D.C.
1978.
Hay,George C., House,Charles E., Sulzer, Richard L., Summary
Report of 1977-1978 Task Force on Crew Workload Federal
Aviation Administration Report No. FAA-ES-78-5.
December, 1978.
as _ Hun Factors
Helm,W. R., Function Description Inventoj
An Empirical Validation Stud!
Test and Evaluation Tool:
U.S. Naval Air Test Center,
Patuxent River, Maryland:
SY-127R-76, July, 1976.
Helm,Wade R., The Application of Computer Aided Evaluative
Techniques to System Test and Evaluation Proceedings of
the Human Factors Society - 21st Annual tleeting, 1977.
Hillier,Frederick S., and Lieberman, Gerald J., Operations
Holden-DayInc.
1967.
Research San Francisco;
HicksThomas G., Wierwille,Walter W., Comparison g& Five
dental Workload Assessment Procedures n A Moving- Base
Driving Simulator Human Factors, 1979, pp 169-181.
Hunter, et al, Modelinq _Ai
raffic Performance Measures.
Volume I.
Message Element Analyses and Dictionaries.
Federal Aviation Administration Report No. FAARD-73-147,I NTIS AD-782 437, July 1974.
Hunter, et al, Simulations Mcdel for New York Air Traffic
Control Communications Federal Aviation Administration
Report No. FAA-RD-74-203 NTIS AD/A-006-426, February,
1975.
-
186
-
Jex,Henry R., ClementWarren F., On Defining and Measuring
Perceptual-Motor Workload in Manual Control Tasks System
STI Working
90250.
Technology, Inc. Hawthorne, CA
Paper 1104-1, July 1977. Tool: An Empirical Validation
Study Patuxent Eiver, Maryland:
U.S. Naval Air Test
Center, SY-127R-76, July 1976.
Kalsbeek,
J.W.H., Do You Believe in
1973, 16, pp 99-104.
Sinus Arrythmia?
Ergonomics,
LauberJohn K., Billings, Charles E., Stevenson, James E.,
Ruffel-Srith, H.P., Cooper, G.E., Simulation Studies, of
Air Transport Operational Problems Proceedings of NASA
Conference: Aircraft Safety and Operating Problems,
NASA SP-416.
October 18-20, 1976.
Hampton, VA.
Laveson, Jack I., NASA Flight Management Research
Proceedings of the Human Factors Society - 21st Annual
Meeting, 1977.
Loftus, Geoffrey R., Dark, Veronica, J., Williams, Diane,
Short-Term Memory Factors in Ground Controller/Pilot
Communication Human Factors, 1979, pp 169-181.
Mental Workload: Its Theory and
Moray, N. (Edit.)
New
York: Plenum Press, 1979.
Measurement
Morgenstern, Bruce, Evaluation of the Airborne Traffic
Situation Display as a Backup Terminal ATC Device
Unpublished Master's Thesis, Department of Aeronautics
and Astronautics, Massachusetts Institute of Technology,
February, 1975.
Mulder, G., Sinus Arrythmia and Mental Workload From Moray
(Edit.) , Mental Workload: Its TIheory and Measurement.
Plenum Press, 1979.
New York:
Murphy, M.E., Coordinated Crew Performance in Commercial
Aircraft Operations Proceedings of the Human Factors
Society
- 21st Annual
Nunnally, J.C.,
1967.
eeting,
1977.
Psychometric Theory New York:
McGraw Hill,
Ogd'en, George D., Levine, Jerrold M., Eisner, Ellen J.,
Measurement of Workload by Secondary Tasks Human Factors,
1979, pp 529-548.
Smit, J., Pilot Workload Analysis Based Upon In-Flight
Physiological Measurements A Task Analysis Methods From
Sheridan and Johannsen (Edit.), d'onitoring Behavior and
Supervisory Control, New York: Plenum Press, 1976.
-
187
-
Smith, H.P. Ruffel-, A Simulator Study of the Interaction of
Pilot Workload with Errors, Vigilance, and Decisions NASA
Tchnical Memorandum 78482, 1979.
Stephens, A.T., Tole J.R., Ephrath, A., Young, L.R., Pilot
Eve Scanning Behavior as an Index of Mental Loading
Proceedings of the 8th Annual Northeast Bioengineering
Conference, MIT, March 27-28, 1980.
Tulga, Kamil M., Dynamic Decision Making in Multi-Task
Supervisory Control:
Comparison of An Optimal Algorithm
to Human Behavior Man-Machine Systems Laboratory,
Department of Mechanical Engineering, dassachusetts-Institute of Technology, September, 1978.
Verplank, William L., The Facilitating Effects of
Uncertainty in Long-Term Manual Control Proceedings of
the Fourteenth Annual Conference on Manual Control,
Massachusetts Institute of Technology, June 1978.
Wherry, Robert J., The Human Operator simultor in th2
Identification of Potential Workload Problems Daring
System Development Proceedings of the Symposium on Man
System Interface: Advances in Workload Study. Air Line
Pilots Association.
Washington, D.C., July 31 and August
1, 1978.
White, R.T. , Task.Analysis Methods: Review and DeveloPgent
of Techniques for Analyzing Mental Workload in MultipleTask Situations McDonnell Douglas, Douglas Aircraft
Company, Eeport No. MDC J5291 September, 1971.
Wiener, Earl L., Controlled Flight Into Terrain Accidents:
System-Induced Errors Human Factors, 1977, pp 171-181.
Wierwille, Walter W., Physiological Measures of Aircrew
Mental Workload Human Factors, 1979, pp 575-593.
Williams, Douglas H., Simpson, Carol A., A Systematic
A proach to Advanced Cockpit Wag.ng Systems krg Air
Transport Operations: Ling Pilot Preferences Proceedings
of NASA Conference: Aircraft Safety and Operating
Problems, Hampton, VA
October 18-20, 1976.
NASA
SP-416
Williges, Robert C., and Wierwille, Walter W., Behavioral
Measures of Aircrew Mental Workload Human Factors, 1979,
pp 549-574.
Yoerger, Dana E., Sheridan, Thomas B., Tulga M. Kamil,
Daryanian, Bahman, Experimentatio-n in Supervisory Control
and Flight Management Progress Report, Man-Machine
Laboratory, Department of Mechanical Engineering,
Massachusetts Institute of Technology, September 1978.
-
188 -
Appendix A
SCEE TOOLS FOR WORKLOAD ANALYSIS
describes several
This section
available for
analytic tools
quantifying cockpit activity and
that are
for charac-
These techniques are useful for
terizing cockpit workload.
analyzing flight procedures in the
design phase,
priori investigations of crew workload.
They
and for a
are particu-
larly useful in that they serve to quantify what has historically been
Moreover,
a very qualitative subject.
these
analytic methodologies focus on the mechanisms through which
the ATC system interacts with cockpit activity.
A.1
Cockpit Activity Timelines
the standard format
Cockpit activity timelines represent
for
the
quantification of
evolved from the
and represent,
crew activity.
cockpit
late 1930's idea of
simply
stated,
Figure A.1
The figure portrays
activity.
Timelines
scientific management
a time and
provides an
100 seconds of cockpit
motion study of
example timeline.
activity during
the localizer intercept phase of a Boeing 707 arrival.
-
189 -
1",
2';i.
2)Se t Bug
3)Turn 130
1)e
1)e
uu
135)Level -of f
-_36) Tr im
7
H500f pmtT
(2500ft)
I TL7
T
So
20 ts
IH200ktsi
3 Select 20HFla
NAVIG. AND
Tune 119,J
PLANNING
TASKS
OP:
83)CL 54,Turn
~
J.2)C
lap201OOZ~InTwr
130
1.4)CL 51 .,Contact
54,Trn
~110
9m
U
d
±Z
8 4Resp., 401 2:sp. Flp4)ep
1320.
Figure A.1:
13.
.
V
1}6o
.
13§0
100 Seconds of a Cockpit Activity Timeline
)
A.1 groups tasks into five
The timeline format of figure
categories.
These are
1.
Heading Tasks
2.
Altitude Tasks
3.
Speed Tasks
4.
Navigation and Planning Tasks
5.
Communications Tasks (Includes both
intercrew and ATC.messages.)
Time into the flight mission is
depicted,
in
seconds,
along
the bottom edge of the figure.
Discrete and continuous tasks are denoted on the timeline
by the shaded boxes.
The length of these boxes corresponds
to the nominal time required
to perform this task.
though placed at some nominal time of occurrence,
relative to one another along the dotted 'task
is,
pilots
have some
degree of
tasks and may choose to
for a
task,
may slide
tracks.'
flexibility in
executing
levels increase.
Tasks
limit
a task
may not
(i.e. task precedence is fixed) and
they may only slide toward increasing mission times.
over,
That
defer the execution initiation time
as workload
slide across one another
Tasks,
may only slide as far as
(the darkened T1 tars).
-
191
-
More-
its associated task
For example,
gory,
in
figure A. 1,
under the heading task cate-
Tasks 10-13 may slide as far as the Task Limit bars at
1340 seconds.
Tasks 10 and 11(21) may not, however,
cross
over,
and thus occur later than,
larly,
the figure shows that Heading category Task--14 and 15
may not slide at all. (22)
slack.--
zero-
-
Tasks 12 and 13.
Tasks 14
Simi-
and 15 are said to have
----
---
Uses of Activity Timelines
Because of the cockpit activity timeline's detail, it
be used in
figure
a variety of ways.
A.1 is
particularly useful
locating tize intervals
where
Notice,
task
First,
in a
format used in
visual sense
where busy periods are
interruption and
for example,
the
multiplicity
from
for
extended or
is
pronounced.
at 1380 seconds on figure A.1 that five
categories of task are occurring simultaneously.
readily apparent
may
the figure how the
communications tend to interrupt
It
is also
intercrew and ATC
other task activity during
this portion of the arrival.
(21)*Tasks 10 and 11 are not
time of occurrence is
on the figure.
Their nominal
earlier than 1300 seconds.
(22)*Tasks may only slide toward increasing time since the
nominal time of occurrence used ' here is equivalent to a
tasks's earliest time of occurrence.
(23)*Timeline data is
either collected
-
192
-
from actual or simu-
The nominal data(23) depicted on the timeline can also be
used to estimate certain workload indices.
A list of these
would include
Percentage of time
1.
busy working on
tasks
2.
No.
3.
Mean task duration
4.
Mean task interarrival
of simultaneous tasks
Cockpit activity timelines are also
of flight procedures to highlight
effort or task multiplicity.
to
Timelines can be constructed
Tirelines
design of workload tolerant,
timeline
loads.
known
can,
thus,
or they may
tasks of spebe used
in the
coordinated crew procedures.
it is also possible to use the cockpit activity
to estimate
physical crew
loads
or mental
crew
Physical crew loads may be estimated by summing the
physical efforts
shown on the timeline.
inputting the
modiJls
figure A.1).
include only the delegated
cific crewmembers.
Finally,
useful in the design
areas of obvious extended
to include all cockpit tasks (as in
be constructed
time
required
Mental
to
lated flight tests,
ing assumptions.
subtasks
efforts can be estimated by
activity data into human
(refer to Hay, et al,
perform the
operator simulation
1978 for descriptions of these
or estimated using a set of operat-
193
-
techniques).
Limitations of Timelines
Cockpit activity
tiwelines,
have
related
to timelines,
noted.
First, tielines,
and
task analytic
limitations
which should
despite their detail,
made accurate to any microscopic standards.
is
be
cannot be
That is, there
a great deal of uncertainty* in the construction of time-
lines which must be recognized.
instance,
Task execution time,
varies with pilot skill, pilot workload,
personality.
Similarly,
there will
may never be
timeline,
performed and still
arbitrariness
in the
A turn,
detail to
not
shown on the
for example,
which activity
all tasks
eye,
Related to this,
which must
Many mental planning tasks occur
There
is
des-
may be described in compliand foot movements,
be more simply described as 'a
seconds duration.
identify
others,
Some tasks
subject to arbitrariness.
cated detail as a series of hand,
or. it may
and pilot
may be included.
The activity timeline is
cribed.
for
be wide variations in
practice from the procedure noted in a timeline.
is
methods
it
turn,'
is
be performed
of several
not possible to
by the
crew.
that are not readily iden-
tifiable and must therefore be omitted from the timeline.
- 194 -
the timeline format of figure A. 1 does not indi-
Finally,
weighting scheme to determine a
common practice to use some
el al,
These techniques seem counter to current theory
(see
and are not used in this report.
1980)
Moray,
Hay,
effort for each task type (see
occupied or on
1978)..
based on either time
workload from timeline data
percentage
is a fairly
It
cate how much effort each task requires. (24)
Summary
The issues
complex subject.
a very
Cockpit workload is
this area are still
at
the forefront of theoretical and experimental research.
The
any quantitative work in
involved in
detail and the nominal information that timelines provide is
a bare necessity,
without which an understanding of the crew
impossible.
workload is
Task Precedence Maps
A.2
a
i.e.
figure
Hold
A.1
by
200
knots.
Holding
solid
-
195
tasks
arrowheads,
with
lines
-
an
are
'easier'
and
attention
Hold
tolerances.
with
continuously
i.e.
task,
intermittent
only
error
large
has
a
and
requires
which
worked
be
must
on
task,
tracking
a
between
distinction
which
task
precision,
strict
task
=
the
make
does
(24)lt
in
indicated
e.g.
that
H200kts
Instrument pilots are trained,
the tasks of
this,
in general,
to prioritize
their multiple task environment.
Because of
there exists an identifiable hierarchy of task prior-
ity and precedence among piloting
certain cockpit events tend to
of other cockpit events.
terminal area,
pilot will
event,
tasks.
Related to this,
cue or trigger the execution
For example. on a descent into the
as. an aircraft approaches 100
slow the
aircraft to 250
'arrive 10000
knots or
feet,
less.
feet,' triggers the execution
the
The
of the
slowing maneuver.
.ask precedence
maps provide a
of the constraints, precedence,
the piloting environment.
dology,
task precedence
crete events that
graphical representation
and triggers which exist in
An analog of the PEET(25) methomaps represent a schedule
take place in the
cockpit.
of dis-
Events,
as
used here, denote the occurrence of a physical event such as
the passage of an IS
message.
marker beacon or the arrival of an ATC
The precedence
map depicts events,
time
of the
events, the uncertainty associated with the event times, and
the precedence which must be maintained between events.
Elements of the Precedence Map
(25)*PERT Planning,
Evaluation,
and Review Technique.
Also referred to as CPM - Critical Path Methods.
See
Hillier and Lieberman, 1967.
-
196 -
A precedence map
cockpit events
constraints.
consists
(see figure A.2)
event paths,
(rectangular boxes),
ing two events is referred to
it
formed
and event
Constraints are indicated by solid, or dotted,
lines connecting two or more events.
raint.
of physical
as a 'hard' precedence const-
indicates that some
between the
A solid line connect-
physical task is
two events,
the
being per-
task duration
denoted by the number enclosed in parentheses.
being
Moreover, an
event to the left must precede an event to its right,
along
connected paths.
A 'soft' precedence constraint (see figure A.3) indicates
that Zvent
1 must precede Event 2,
Event 2 at which
time after
no definite
Event 1 must occur,
physical task.separates the two.
and Event 1,
but there is
in figure A.3b,
since no
The time between Event 2
must be greater than or equal
to zero seccnds.
Associated with each
event is an earliest
time at which
the event can occur and a latest time at which the event can
occur.
These values are indicated by the left and rightmost
enclosed
numbers,
For example,
which the
in parentheses beneath the
for Event 1, figure A.3a,
turn maneuver can begin
mission and the latest time is
between
these two
slack.'
Event slack
numbers
event boxes.
the earliest time at
is 120 seconds
323 seconds.
is referred
to
The difference
as the
represents the 'uncertainty in
-
19 7
into the
'event
the time
I
(N
/
/
/
/
/
-
--
-
.
-
--
-
-
-
aContac
-
tower"
/
(1387,1600)
0
.
Stop
turn
(1357,1357)
.S tart
intcpt
Iturn -(1387,1387)
Stop
turn
(15)
..
..
.. ..
...Q o .... \---
-
1402,457)
E4
E~ ~
~
~
~
a -- -- -- .... -- ... -&
-MN
- - - '- - -- '- -- -- . . . ..
.
eg in
7
-echan.
te le I j t
-
(1387,1600)
Figure A.2:
A Task Precedence Map
P6
E
(120,323)
a)
(127,330)
Hard Precedence Constraint
2
1
BeginBegin
Pitch
i
Mnv r.
(120,330)
(120,330)
b)
Soft Precedence Constraint
Figure A.3:
Precedence Constraints
-
199
-
of occurrence of
It
each event.
path along which the slack is
should be
noted that the
always zero is
referred to as
the critical path.
Contructing a Task Pregedence Map
A task precedence
map may be constructed
by following a
simple four step procedure.
1.
Break down
events
the task
into a
componcnts and
desired
level
cockpit
of detail
of
events.
2. Set the nominal time of
.
task or event occur-
rence, on the activity tineline,
to the ear-
liest time for an event.
3.
Sweep forward through the
events have
raints
timeline until all
been placed on the
indicated,
map,
const-
and
earliest
times
final
scenario
event,
assigned.
4.
Beginning
with
the
sweep backwards through the timeline, setting
the 'latest
time'
time'
to the
smallest 'earliest
imposed by the constraints.
-
200 -
An example using figure A.2 will illustrate this process.
Cn the
activity tizeline,
Event D22)
nominally
begins at 1387 seconds
At approximately this same time,
sion.
the intercept turn,
miles outside of
(associated with
the intercept
into the mis-
but definitely after
the aircraft arrives at the localizer,
the outer marker.
The
5
activity timeline
indicates, also, that the intercept turn requires 15 seconds
and so event D23 occurs at 1402 seconds
earliest.
ATC
arrives at,
but not before,
will
be told
procedure dictates
(1387 + 15),
that when
the aircraft
the 5 mile point,
to contact
the tower
at the
the aircraft
(Event C10).
Flight
procedure further dictates that
the Mechanical checklist be
initiated upon intercepting
localizer.
is
the
Thus Event E4,
constrained to occur at or after Event A6,
which in this
case occurs simultaneously with Event A7.
Event C10 is assigned an 'earliest time' of 1387 since it
is
constrained
earlier than event
to occur no
larly,
Event E4 may then be
1387.
Should two
be the
assigned an 'earliest
Simitime'
of
or more precedence constraints
affect a
then the 'earliest time'
assigned
single event, C10 perhaps,
should
A7.
largest
value
imposed by
the
appropriate
constraints.
'Latest times'
the final event.
by sweeping
For instance,
Event A8 (not shown)
'latest time' of 1457.
the 'latest
time'
backwards from
are assigned
has a
Event A7 mu'st therefore be assigned
of
-
201
-
Minimum (1457-70 or 1600 seconds as imposed by Event C10)
This process
continues backward until
been assigned a 'latest time'.
the first
Note,
event has
on figure A.2,
along the double line the slack is always zero.
double-lined
path indicates
the location
Thus,
that
the
of the
critical
uses in
workload
path.
Uses of Precedence laps
Precedence
maps have
several obvious
analysis or flight procedure design.
use of the slack values.
be in the
the
The most obvious may
time slideability
of tasks
Thus, for each cockpit task,
and placed
on the cockpit
These values indicate
defined
on the
timeline.
a Task Limit may be identified
activity timeline.
. Figure A.1
shows the presence of Task Limits by darkened TL bars, where
these bars are located at
the 'latest
time'
associated with
specific events.
The value of the slack,
itself,
may indicate the poten-
tial for high workload situations, or the workload tolerance,
of a set of flight procedures.
large,
tasks associated
for long periods
duced.
with these events may
should tasks of higher
Similarly,
tasks are not
That is, if slack values are
if
priority be intro-
mean slack values are
readily deferrable,
-
202
-
be deferred
and the
small,
then
introduction of
sources may lead to poten-
high priority tasks by exogenous
tially high workload situations.
The precedence map also highlights
triggers by indicating which
Thus,
in figure A.2,
point triggers
events constrain other events.
the aircraft's
the radio frequency
other instances
the existence of task
one event
arrival at the 5 mile
command to
may trigger
occur.
a long
sequence of
other events and their associated task to occur.
some cases,
be appropriate to
It may, in
change procedure if
sequence becomes inordinately complex
In
the task
with very small aver-
age slack values.
Finally,
critical
the precedence map indicates
event path.
path is
The critical
constrains all other paths and events.
cal path is changed the 'earliest'
other paths will be affected.
constraints.
lengthened,
or events
If
that path which
Thus, if the criti-
and 'latest'
Moreover,
generally indicates which path or
the dominant
the presence of a
times on all
the critical path
set of events is imposing
path is
the critical
along the critical modified
tructuring flight procedures or
ATC procedure,
thus
by res-
the nominal
workload will be modified.
Summarizing,
tool
pilot
task precedence maps
for evaluating
or crew
cockpit procedure
They
workload.
are
are a useful analytic
in
derived from
activity timeline data but are different in
-
2023
-
the context
that
of
cockpit
1.
They
identify a
events which
critical path
of
other
constrains all
cockpit events.
2.
They
highlight
formal
precedence
constraints.
3.
They indicate which system elements
are triggering other cockpit events
to occur.
The precedence mAp is generally most useful when constructed
with a more
macroscopic level of detail as
microscopic detail with which
is
compared to the
activity timelines are gener-
ally constructed.
It
suggested that
these two
tools
should be used in a
complementary manner in workload analy-
sis.
Stochastic Analysis of ATC Communications
A.3
It
has been shown that ATC communications are a source of
loading to an air transport crew.
This loading acts through
two types of mechanisms.
1.
Explicit mechanisms -
ATC messages
in the form of advisories or clearances that cause the
pilot to exe-
cute some set of physical or mental
planning tasks.
-
204
-
2.
Implicit mechanisms -
implicitly
interrupt and
Some messages,
the crew.
ere
A2C messages
weather
warnings
arouse anxieties
distract
e.g.
sev-
can
also
and confusion
in
the cockpit.
The extent to which explicit
be
determined
precedence maps.
from
mechanisms affect the crew can
timelines
cockpit activity
Implicit loading, however,
of the random occurrences of ATC messages.
and
task
is the result
It is useful to
examine some of the features of this stochastic process.
Each message
crew.
distracts or diverts
characterizes
'message occurrence,'
of messages or
(1975)
the
magnitude of
in
a given time interval,
mean message rate can
this
the mean no.
be computed.
Dunlay
has exairined this in some detail.
The most comprehensive
treatment
of the statistical
perties of ATC communications was performed by Hunter,
(1974)
is
Given a probability distribution for
interruption process.
tion.
of the
The probability that a message will occur, then,
parameter which
one
the attention
in
the development of
proet al
an ATC communications simula-
Hunter defined three groups of data that characterize
all ATC message activity.
These are
-
205
-
Communication transaction
1.
dialogue between
(CT)
- A
a controller
and
one aircraft.
-
(TR)
Transmission
2.
When a micro-
phone is keyed.
- One phrase
Message element (ME)
3.
of which
or message
or many,
none,
sector an aircraft
For every CT there
typically participates in several CTs.
and there may be
one or more TRs,
may be
be
a transmission.
in one ATC
and A.5 show,
As figure A.4
in
there may
several message
elements in each transmission.
CT/aircraft,
In addition,
TR/CT,
the
random variable.
lated
over,
length of time of each
transmission is
a
Hunter created 'dictionaries' which tabulisted
elements and
all message
transmissions in
and ME/TP are all random variables.
which these elements
the proportion
occurred. (26)
of
More-
he identified probability distributions that described
CT/aircraft,
craft in
found the
TR/CT,
and TIBE/TR,
For
a control sector.
given the number of airthe New York
area Hunter
following distributions to representative
of the
data
(26)*Hunter used data from the
areas,
-
New York and Houston control
206 -
Single Aircraft
(
PerforrL ce
Communicptions
time in' seconds
AA
i.
S-
miwn
w.
n
'I'.'
U
La
f~2
F-
Communications Transaction
(time length varies)
'I
Aircraft Initial CT on Entering Sector
Aircraft Final CT on Departing Sector
Composite
L.
-g
~--
Communications
A
C3]time
Performance
~IUWUUE~4I
-
in
1
scn
A
time in seconds
Figure A.4:
WE
1W9 U V9 WE
A
-
Air Traffic Communications Performance
-
~
a
2
kin
/
Aam
en=ine
w
A~ .~
nm
-
Single CT
Transmission 2 (Pilot)
I
1...-
.
*
(Transmission)
*
\1
0 1.
Pause
Altitude Con. Altitude
.:
TR Time in Seconds
Two Message Elements
Figure A.5:
)
I
< 2 sec.
Decomposition of a Connunications Transaction
.
.1
*
NI\
I
Trins. 4 (Pilot)
Transmission 3 (Cont.)
CT' Time in Seconds
Single TR
/
0 1 r-t L ML P
rn
Transact
inn)
(Communica tions Transactinn)
Transmission I (Cant.) I
E L N-. k
en
mm
.
.3
*J
CT/aircraft
-
Doubly
shifted,
negataive
binomial
distribution
TE/CT - Truncated,
negative binomial distribution
Time/TR - Gamma Distribution
It
is
suggested that
in workload
tions to Hunter's techniques should
simulation approximabe used in constructing
doreover,
it is
a realistic ATC communications
scenario.
suggested that in the workload
analysis of a mission phase,
the statistical parameters that
describe the ATC communica-
tions provide an indication of the magnitude of the implicit
loads due tc messages.
-
209
-
Appendix B
FIGURES FOR THE
ANALYSIS OF
210 -
A BOSTON AERIVAL
COCKPIT ACTIVITY TIMELINES -
Task #10
Duration=10 sec.
Task #11
Duration=l
LEGEND
Task Slide Limit for
Task #11
sec.
Tills
TL1O
Head ing
-..0
Tasks
Task category
Task Sliding Track.
Tasks may slide from their nominal time (shown)
to right only.
Tasks retain precedence depicted, i.e. Task 10 may
slide as far as TL 10, but it may not cross over Task 11.
Al t itude
H2000 fpm
Tasks
Indicates a Hold type task, i.e. Hold
2000 feet per minute descent rate.
Time, in seconds, into mission.
Hold type tasks require only intermittent
attention from crew.
80o
8010
100
])Set Heading
HEADING
2)Track 107 Radial
TASKS
)Set Alert
2) Pi tch Down
3)Trim
ALTITUDE
H3000ffm
I(FL330
TASKS
SPEED
1)Throttles -Idle
H .82M4
TASKS
L-a-
-
-mmmmmmmmmmmmmmmmmmmmm
_
m--
-
-
-
-
-
- -
m
m
NAVIG. AND
PLANNING
TASKS
COMMUNIC.
1)CL54 Mintain FL220
TASKS
J2)
Res ponse ,CL 54
20
Figure B.l:
40
60
80
Cockpit Activity Timeline, ILS 4R Arrival
100
HEAD ING
~t
'fll~
Ra ial
2)Track 107
Uf~
AL TITUDE
(FL300) H3000f m
--
TAS KS
-mm -
=m
mo
a
-m
-m "M
-m
"M
-------------
-m
-
-m
m
-
mm
-
-o
-
-m
m
-
-
-
------------------------------------------------------------------------------------------
-
-m
SPEED
H .82MI
TAS KS
1;
NAVIG.
X
- -- - -- - -- - -
AND
---------------------------------------------------------------------------
----- --------
PLANNING
TAS KS
AA388,Maintain
FL220
COMMUNI C.
TAS KS
3)BOS Cent.,
AA.8. .-----
120
10
60
20
180
P2
)
4
HEADING
5)Track 118 Radial
TASKS
TL 1-8
9)Set Altimeters
ALTITUDE
to field
H2000;
(FLl90)
TASKS
TL1
SPEED
H320 kts
TASKS
NAVIG. AND
PLANNING
TASKS
-
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
-
m
COMMUN IC.
TASKS
m
m .
420
m
m
m
-
m
m
440
m
m
m
m
m
m
m
m
460
m
m
m
m
480
-
500
P5
DING
-i
5) Track 118 Rad ialI
KS
10) Pi tch Up
I TUDE
.
1 )Trim to 500f
H2000fpm
(16000
KS
ED
H320kts
KS
IG. AND
NNING
KS
Smmmem
MUNIC.
m
m
m
mmmmmmmmmmmmmmmmmmmmmmmmmmm
NN
MN
E
KS
mm
mw
m
1)AA388 clear 8000
cross GDN 14000
12) Response,
AA388
540
m
560
m
I 3)P2:1000ft
Above
Above
-580
600
P6
14)Set Alert
115) P itch Down
12)Pitch Level
ALTITUDE
13)Trim
TASKS
m mm m m m m
16) Trim
(14000ft)
AIO NNN OWMNMEAm m aimm mmmm
mmmmmummmum
m mm mmmm mmlMImmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm
mmmmmm
H320kts
m m
M
AMImM
IAMIm
(slowing
TL2
m m m mAmmm
mimm
mn
MEANm MEAm mummmmmm
2)Set CDI 111
degrees
Moms mmmm
m
620
mN mE Msm
m
mN m
m1
mmm
649
m
m
660
(i)
"m9m m
m
m
m
m
680
m
m
m
m
m
m,
700
a
TASKS
3)Slowm t o2500
SPEE
TASKS
TAS KS
5,4
720IG
,AND.
2:10
7,07,0
P8
P9
'6
A(
-
M
-
am
-ny
m
mmm
wmm
mom amm
ALTITUDE
IH2Obo pm
TASKS
M
mm
~~~~~
"M
=
o
o=9 mo
H250 or less
-=moo
m
Mm M
m
ow
M
mUW
mWm
-
Mm
in
L22)P2:1000ft
Roger
Above
23)Ck
i
24)Leaving
]st
6)k ist
MANJO,CL 54
9A
94o
i
960
.
80
.
.
4
PO
HEAD ING
H160deg
TASKS
160deg.
TL 14-1
24)Level off 6000ft
25)Trim
ALTITUDE
HR6000ift
TASKS
4)Throttles
to 1.6EPR
SPEED
H250kts or less
______
TASKS
m
NAVIG.
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
m
AND
PLANNING
TASKS
2AA38 ,Depart
MANJO 160,6000ft
COMMUNIC.
TASKS
-7)! nitIial
Cal 1-up, AA388
5
9Q20
)q40
30)CL 54,Turb.
Advisory
29)Resp.,
AA388
1060
31Resp.,
CL 54
,
100
(00
Pill
H160de
27)Pitch
-
---
-
T26)Set Alert
ALTITUDE
Down
H2000fpm
28)Trim to 2000fpm
TASKS
.,
TL
m
m
ii§mm~
imomm
mmm
on"
mk-
2 j.3.
)Throttles
IdleH250 or less
H220kts
IO)Tune 110.3
OW
iim
COMMUNIC.
m
mm
m
imm
2m
m
m
m Oremmm
m
mm
m
mamn
m
m
m
mm
mmi
32)CL 54 ,
Clear 3000
TASKS
34) P2:1000-
33)S Resp.1,
CL 54
mom~~*
u1g0Q
11,40
M=
.Gsmom
-
JiWi,
inm m m m M~
u680 ,
,
,
0
Above
nr~uuLlm
m
1290
P12
Im
TASKS
NAVIG. AND
2)OSeat Belt
n
mSi
mm
PLANNING
TASKS
36"CL
COMMUNI C.
2
54,
f
clear
-
TASKS
t35) Ck li st
#12.
1220
1240
13)Res p.
CL S
1260
1280
1300
P13
LT
TL I O
HEADNG)Tu
130
I
16)Set Bug
4)Set Bug
7) Intercpt
08
T
Turn
TASKS
T5-m
m-I
-f7
f'm
-v
36)Trim
ALTITUDEH5fm
TASKS
T2
TL
-
SPEED.)hote
TASKS
I3Slect
eclect 30 Fa
(250 1f
55e
.e
,1ue1
20 F-
ec
30FlS
TASKS
COMMUNIC
3 CL 5 ,Turn
130
TASKSResp
42)C
Pl:
F1ap2
41)P2: Flap
~~CL
5
3
n
s
44)CL
4,Contact
,
47)P2:1ap3
CL 54
206 CL 54
54
1320
20
340
.
0
46)PI :Flap
0
130
1L
0
P1 4
-
HEADING
/
£
4,
18Tack Localizer
TASKS
we
-
TL
35, 36)Td
ALTITUDE
3 )Pitch Cag
TASKS
)T r. i- 40)Track G1 des o e
T
r
-
m
-
m
-
m
------
m
-
SPEED
- -~am
-
m
-
m
-
-
01
---
M"
H
40kts
TASKS
mm
NAVIG.
m
WN m IIIIII01
an
IIIIm
AND
PLANNING
m
m mmw
m
TASKS
eso to
51)DL s411
52)Resp.,
COMMUNIC.
5)Re
TASKS
3 ..-
id
-... L-- (PI
Ckist
clear
toff
DL 411
53)PI:Gear,
Fla
40
55)P 2:Pumps on
Press normi
---
and P2)
ima iap 40 LAm
1
1420
.
1440
18)20 a----54TP2:Gea r ,56)P2:Spe
ed
Bra ke Fow' d
1460
1480
15007
P15
7£)
"C
I*
P16
-
228
-
Ir
(
fASK PRECEDENCE
MAP
-
LEGEND
(20)
(60)
Event Box
Event No.
B
Event Description
vent Path
Indicator
Latest time of occurrence
Earliest time of event
occurrence
- Position Event Path
B
0
0
- Altitude Event Path
''Hard precedence constraint.
Al occurs 20 seconds before A2.
- Communications Event Path
A2
- Manual
Control Event Path
- Planning Event Path
'Soft' precedence constraint.
B3 may occur at any time
after event A2.
(360)
(300)
Grisy XN
(300 ,300)
B3
--
-
(300,937)
(0,637)'
(80)
..
Leave
1800oft
\(300,937)
380,1017)
C
I
/_
0
4--
'
I
_
'SC
I
I
I
/
DI
~~Begin
B
-Leave
22000ft
I
.........
..
..
..
..
.
. e.
Descent
(0,466)
E2
-----------------
E
180ooft
Cklist
(380,1600)
(0,1600)
Pd
Figure B.2:
Task P:
Iff1 I 0
)dence
Map, IL71S 4R Arrival)
.
1
t
I
0
GA)
-
-
-
-
-
B5
Arr ive
liOQOftI
---V
I
/
(61O
7110
/
-----
LeveI
-of
(430,826)
fDescent
40826 )to 80QQ
(660,826)
\
DA__
evel-
- 00
0
(760 ,1
250k~t
(660,1126)
(660,1600)
P2
--KD
-B
LeaveB
6000ft
(1140,1140)%-
--
-
-
-
-
-
Bg In
-Descent
"-
-
'-
-
--
-D
[to 3000f
(994,1240)
(0140,1240)
P3
r~T1
~
KI
/
(9Z~L
(9zct'o9z~)
/
/
-
i 408n)
&9
'~
S
_______
--
-
~
A J
B ig
0----------------.
-----------
-In'
m
m
m
m
m
m
m
m
mgo
-m
-A
m
m
m
m
--
m
m
m
-----0
c
(1340,134o)
(1300,1300)'
134Rf
L-E
E
-
i
I JU-Iilrn
i
_(024.1340O)
E(280jl32L6)
LLLU...JJ
--
-----
-
---
----
P5
40
I
/
/
/
-
/
/
/
/
/
@11~~~~~~~
_
-
/
-
__
_.- "ontacC
tower"
/
(1387,1600)
(15)
,1357)
(1387,1387)
D
457)
(1387,1600)
(1457,1457)
4E
E
P7
)-
R
q
Appendix C
EXPERIENCE QUESTICNNAIRE
-
237
-
SUBJ NO. I
PILOT EXPERIENCE QUESTIONNAIRE
SubJect Name
Part 121 or 135 Experience?
Yes
TEST NO.a
No
Total No. of Hours
Total No. of Hours under Simulated or Actual'IFR
RECENT EXPERIENCE
Last 30 Days
Lost 6 Months
Hours Total
Hours IFR
(Simulated or
Actual)
No. of Instrument
Approaches
Estimated No. of
Non-Precis ion
Approaches
Do you, on a regular basis, fly into the following
terminals (check as many as necessary):
Atlanta
Los Angeles
San Francisco
New York
Washington, D.C.
Chicago
Type of aircraft you fly most often (list one, or at most, two)
I-
-
238
-