Content uploaded by Bernardus F Willems
Author content
All content in this area was uploaded by Bernardus F Willems
Content may be subject to copyright.
THE EFFECT OF AIR TRAFFIC INCREASE ON CONTROLLER WORKLOAD
Sehchang Hah, Ph.D., Ben Willems, M.A. &
Randy Phillips, Supervisory Air Traffic Control Specialist*
Human Factors Group-Atlantic City
Federal Aviation Administration (FAA)
Atlantic City International Airport, New Jersey
The Federal Aviation Administration (FAA) has been increasing the National Airspace System (NAS)
capacity to accommodate the predicted rapid growth of air traffic. One method to increase the capacity is
reducing air traffic controller workload so that they can handle more air traffic. It is crucial to measure
the impact of the increasing future air traffic on controller workload. Our experimental data show a linear
relationship between the number of aircraft in the en route center sector and controllers’ perceived
workload. Based on the extensive range of aircraft count from 14 to 38 in the experiment, we can predict
en route center controllers working as a team of Radar and Data controllers with the automation tools
available in the our experiment could handle up to about 28 aircraft. This is 33% more than the 21
aircraft that en route center controllers typically handle in a busy sector.
The Federal Aviation Administration (FAA) predicted
that air traffic will grow substantially in the coming
years (FAA, 2006). To accommodate it, the FAA has
planned to increase the National Airspace System (NAS)
capacity by building new runways, modernizing
hardware and software, and modifying existing
procedures. With improved workstations, controllers
can handle more aircraft, which will increase the NAS
capacity. Recently we tested new concepts to improve
controllers’ workstation with various traffic levels. In
this paper we report the effect of increasing traffic levels
on controllers’ perceived workload.
Controllers “coordinate the movement of air traffic to
make certain that planes stay a safe distance apart. Their
immediate concern is safety, but controllers also must
direct planes efficiently to minimize delays.”
(Department of Labor, 2006). To achieve their goal,
controllers monitor situations, resolve aircraft conflicts,
manage air traffic sequences, route or plan flights, assess
weather impact, and manage sector/position resources
(Alexander, Alley, Ammerman, Hostetler & Jones,
1998). They perform these tasks concurrently and
expeditiously in their own responsible airspace called a
sector. They also convert the information about aircraft
into three-dimensional space and sequence them to
safely leave the sector within the constraints of written
agreements. Automation and decision aid tools such as
conflict probe (User Request Evaluation Tool [URET])
and traffic flow advisory (Traffic Management Advisory
[TMA]) are available for controllers’ use. Most of the
information about aircraft for controllers to perform their
task is presented on the controller’s monitor screen.
Aircraft are displayed with a diamond symbol and have
an attachment called a data block furnished with the
critical information about the aircraft. Data block
formats are domain specific and thus different in a
Terminal Approach Radar Control (TRACON) and Air
Route Traffic Control Center (ARTCC).
Since our experiments were run in the ARTCC
environment that uses the Display System Replacement
(DSR), we used the data block format shown in Figure 1.
(See Table 1 for the description of the data block). In
our experiment, the typical visual angle for the aircraft
and data block with the shortest leader line was about
3.3 horizontal and 1.2 vertical degrees respectively
assuming controllers sat about 24 inches away from the
monitor. The lengths of the sector were about 32
degrees vertically and 28 degrees horizontally (Figure
1). From these visual-angle values, it is easy to imagine
the cluttering effect from the increasing number of
aircraft on the display.
Controllers need to direct the arrival aircraft to form
traffic flows and hand off them to the next sector
controllers (Figure 2). As the traffic volume increases,
their data blocks are likely to overlap, and controllers
offset them manually to read the information on them.
Thus, with more aircraft, controllers need to perform
more complex perceptual and motor tasks. This also
increases the complexity of high-level mental tasks such
as memorization and decision making which can
increase workload.
* Now with Cleveland Air Route Traffic Control Center
(ARTCC).
*This work is not subject to U.S. copyright restrictions
PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006 50
Figure 1. Data block. Controllers sometimes use the
temporary fourth text-line to display
temporary heading, speed, or free text.
Figure 2. Airspace with the sector located in the middle
(shown as shaded).
Table 1. Data block and other display elements about USA639 in Figure 1.
Display Elements Term Description
Diamond symbol Target / Aircraft This symbol changes into a triangle if the aircraft deviates
from the flight-plan path on the radar
USA639 Call Sign / Aircraft
Identification (AID)
Controllers usually refer to it when communicating with
pilots and other controllers.
Solid triangle symbol on the
first line
Data-link Symbol
If an aircraft is not data-link equipped, this symbol will not
be shown. Seventy percent of the aircraft in the experiment
were data-link equipped.
310 Altitude Assigned altitude: 31,000 ft.
C Altitude Profile
Indicator
“C” stands for cruise or level altitude.
163 Computer
Identification
Number of the
Aircraft (CID)
Controllers usually use it for keyboard entry pertaining to
that aircraft.
G Destination Symbol This G stands for the destination airport, Genera Airport.
434 Ground Speed Ground speed in knots.
Solid line from the aircraft
to the data block
Leader Line Controllers can adjust the leader line to one of three lengths.
Broken line attached to the
aircraft
History Trail History of the aircraft positions.
Solid line attached to the
aircraft in the opposite
direction to the broken line
Vector Line This line shows aircraft’s heading and projected position.
64.4 Continuous Range
Read-Out (CRR)
This shows the distance between the aircraft and a pre-
determined fix a controller can choose. A fix can be
waypoint, airport, etc., on the display.
CHIGO
INDIN
LINCO
BUTTE
DESMN
TOPK
A
IND
KANCY WHEEL
J26
J7
9
J12
J13
J79
J79
J23
J23
J23
J30
J1
J13
J1
J1
3
J30
J12
J26
J2
6
J22
J2
2
J21
J100
J100
J100
J15
J21
J1
2
J1
2
Sector 08
FL230
-600
120.08
OHO
GEND2
N
J21
J2
5
J23 IOW CIN
SPL
Cross CHIGO @ 230
Cross DARIO @ 230
DARIO
*This work is not subject to U.S. copyright restrictions
PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006 51
The number of aircraft in the sector is a major factor that
contributes to air traffic control complexity. Some
researchers did not identify it as a separate factor but
used factors that were created by aircraft such as
dynamic density. For instance, RTCA defined dynamic
density as “traffic density, complexity of flow, and
separation standards.” (RTCA, 1995). RTCA suggested
that for a short-term environment such as a sector,
dynamic density could be used to measure air traffic
control difficulty.
Most of the research on the effect of dynamic density or
air traffic control complexity on workload has been
based on interviews, surveys, or observers’ ratings, and
not on controllers’ direct ratings while controlling
traffic. (Note: There are numerous reports on this topic.
Review papers by Mogford, Guttman, Morrow, &
Kopakdekar [1995] and Hilburn [2004] have an
extensive list of them.)
Recently, Lee (2005) argued that the relationship
between aircraft count ranging from 6 to 26 and
workload was nonlinear. His participant controllers
rated their workload while controlling air traffic. He
reported that an S Model described it better than a Linear
Model, because the proportion of variance explained
(R2) by the S Model was larger than the one by Linear
Model. Since he did not statistically test the R2
difference, however, we do not know if the difference
was significant. His results were also based on only
three controllers’ data.
We examined the effect of air traffic volume on
controllers’ perceived workload using an extensive range
of aircraft count in the sector from 14 to 38. Based on
the results from the high-end aircraft count, we also
could predict controller workload for the future
increased air traffic volume.
METHOD
Participants
Sixteen full-performance level controllers volunteered in
our experiment. They worked as a team of Radar (R)-
side and Data (D)-side positions. Our participants came
from some of the busiest ARTCCs. D-side controllers
assisted R-side controllers.
Due to computer problems during experimental runs for
the first two teams, we used data of the remaining six
teams for analysis. We also had 6 pilots, and each of
them handled multiple aircraft.
Equipment and Materials
We used two high-resolution 20x20-inch Barco LCD
monitors (2,048 by 2,048 pixels), one for each position.
An in-house real-time simulator emulated the Display
System Replacement (DSR). The R-side had a TMA list
as part of the Center TRACON Automation System
(CTAS) and Controller Pilot Data Link Communication
(CPDLC) Build 1A R-side interface. The D-side had
Computer Readout Device (CRD), URET windows, and
CPDLC Build 1A D-side interface. The FAA William J.
Hughes Technical Center Target Generation Facility
created aircraft target data.
For workload ratings, we used Workload Assessment
Keypad (WAK) (Stein, 1985) that was a 4.25 x 8.5 inch
instrument box located between the monitor and the
keyboard and within an easy reach of the controller.
Airspace
Our participant controllers used a generic air space that
was easy to learn in a short time (Figure 2). They
controlled traffic in a high altitude sector that metered
three streams of traffic to low altitude sectors. Seventy
percent of the aircraft were data-link equipped.
They used 1,000 ft as the vertical separation minimum
and 3 miles for the lateral separation minimum that are
similar to NAS separation standards in 2010 (RTCA,
2002). Currently the lateral separation minimum is 5
miles.
Procedure
Each controller team participated in three experimental
runs that used three different scenarios loaded with 21,
27, or 35 aircraft. However, the number of aircraft in
each run fluctuated to some extent because of the
dynamic nature of air traffic control.
We instructed controllers to press a button from 1 to 10
corresponding to their workload rating: “At the low end
of the scale (1 or 2), your workload is low - you can
accomplish everything easily. As the numbers increase,
your workload is getting higher. Numbers 3, 4, and 5
represent the increasing levels of moderate workload
where the chance of error is still low but steadily
increasing. Numbers 6, 7, and 8 reflect relatively high
workload where there is some chance of making errors.
At the high end of the scale are numbers 9 and 10, which
represent a very high workload, where it is likely that
you will have to leave some tasks unfinished.”
*This work is not subject to U.S. copyright restrictions
PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006 52
RESULTS
There was a high correlation between R-side and D-side
workload ratings (r = .79). The D-side controllers’ role
was to assist R-side controllers. Thus, we analyzed only
R-side workload ratings in this paper. There was a
linear relationship between the number of aircraft and R-
side controller’s workload ratings (t = 18.75, p < 0.01,
R2 = 0.52) (Figure 3). The regression equation was
Workload Rating = 0.306 x Number of Aircraft - 3.373.
For both Quadratic and Cubic Models, R2s were 0.53.
But there was no statistical difference between Quadratic
and Linear Models (F1,326 = 1.83, p >.05). Other models
(Logarithm, Inverse, Compound, Power, S, Growth,
Exponent, Logistic) explained less variance than Linear
Model. For the aircraft count ranging from 6 to 26 that
corresponded to Lee’s aircraft count, there was also the
linear relationship (t211=2.63). However, the Linear
Model had small R2 (.03). Other models also showed the
similarly negligible size of R2.
There were large team differences in workload ratings as
shown in Figure 4. Surprisingly some R-side controllers
(Teams 1, 4, and 6) did not rate their workloads high
even when they handled more than 30 aircraft in the
sector and had difficulty in controlling traffic. (Note:
The reason why ratings of 10 appeared in the scatter plot
[Figure 3] but not in the line graph [Figure 4] is that in
the line graph, ratings were averaged.)
Number of Aircraft
40353025201510
Workload Rating
11
10
9
8
7
6
5
4
3
2
1
0
Figure 3. Scatter plot to show the linear relationship between the number of aircraft and R-side workload ratings.
Number of Aircraft
38363432302826242220181614
Workload Rating
10
9
8
7
6
5
4
3
2
1
0
Team
1
2
3
4
5
6
Figure 4. R-side workload ratings of individual teams.
*This work is not subject to U.S. copyright restrictions
PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006 53
DISCUSSION
Our data showed that controller workload had a
significant linear relationship with the number of aircraft
in the sector, and this relationship described about 52%
of their workload-rating variance. Even though the
quadratic model had a slightly larger R2, .53, there was
no significant difference between them. Thus, for the
sake of parsimony we consider the relationship as linear.
Other models including the S Model did not describe the
data as well as the Linear Model.
The aircraft count has been very robust in predicting
about half of the controllers’ workload variance: 52% by
us, 53% by Hurst and Rose (1978) (quoted by
Eurocontrol, 1998), and 60% by Stein (1985). The
average of Lee’s Linear Model R2’s for three controllers
(0.27, 0.54, and 0.77) was .53, which was not much
different from our .52. This is very intriguing because
all these experimental results showed similar R2’s in
spite of their differences in air traffic levels,
experimental setups, and simulator configurations.
Our extensive range of the number of aircraft, from 14 to
38 aircraft, enabled us to predict controller workload for
the future traffic level. Given moderate workload ratings
of “5” in our experiment with low probability of errors,
we can predict that controllers could handle up to about
28 aircraft using the DSR with CPDLC, TMA, and
URET tools (Figure 3). This 28 aircraft is about 33%
more than the 21 aircraft that ARTCC controllers
typically handle currently where they do not have the
three technologies optimized the way we accomplished
it in this experiment.
ACKNOWLEDGEMENTS
The authors wish to thank our programmers for their
computer programming support, Dr. Earl Stein for his
comments and suggestions, Denisse Villa for her
secretarial support, and Linda Johnson for her editing
support. Especially, the authors express their gratitude
to controllers who volunteered to participate in the
experiment.
REFERENCES
Alexander, J. R., Alley, V. L., Ammerman, C. M.,
Hostetler, C. M., & Jones, G. W. (1988). FAA air
traffic control operations concepts, Volume III: ISSS
en route controllers (DOT/FAA/AP-87-01).
Washington, D.C.: Federal Aviation Administration:
Department of Transportation.
Department of Labor. (2006). Air traffic controllers.
Retrieved February 1, 2006, from
http://www.bls.gov/oco/ocos108.htm.
Federal Aviation Administration (FAA). (2006). Flight
Plan 2006-2010. Retrieved February 1, 2006, from
http://www.faa.gov/about/plans_reports/media/flight
_plan_2006.pdf.
Hilburn, B. (2004). Cognitive complexity in air traffic
control - A Literature review (EEC Note No. 04/04).
France: Eurocontrol Experimental Center.
Hurst, M. W. & Rose, R. M. (1978). Objective job
difficulty, behavioral response, and sector
characteristics in air route traffic control centers.
Ergonomics, 21, 697-708.
Lee, P. U. (2005). A non-linear relationship between
controller workload and traffic count. In Proceedings
of the Human Factors Society 49th Annual Meeting
(pp. 1129-1133). Santa Monica, CA: Human Factors
and Ergonomics Society.
Mogford, R. M., Guttman, J.A., Morrow, S. L., &
Kopardekar, P. (1995). The complexity construct in
air traffic control: A review and synthesis of the
literature (DOT/FAA/CT-TN95/22). Atlantic City
International Airport, NJ: Federal Aviation
Administration, William J. Hughes Technical Center.
RTCA, Inc. (1995, October 26). Executive summary of
the final report of RTCA Task Force 3: Free Flight
implementation. Washington, DC.
RTCA, Inc. (2002). National Airspace System.
Concept of operations and vision for the future.
Washington, DC.
Stein, E. (1985). Air traffic controller workload: An
examination of workload probe. (DOT/FAA/CT-
TN8424). Atlantic City International Airport, NJ:
Federal Aviation Administration, William J. Hughes
Technical Center.
*This work is not subject to U.S. copyright restrictions
PROCEEDINGS of the HUMAN FACTORS AND ERGONOMICS SOCIETY 50th ANNUAL MEETING—2006 54
