COLLABORATIVE DECISION
MAKING FOR AIR TRAFFIC
MANAGEMENT: A
PRELIMINARY ASSESSMENT
Prepared by
NEXTOR
The National Center of Excellence for
Aviation Operations Research
The NEXTOR Team
University of Maryland: Michael Ball, Robert
Hoffman, Tasha Inniss,Thomas Vossen,
Chien-Yu Chen, Daniel Darr, Joseph
Previte
MIT: Amedeo Odoni, William Hall, Alp
Muharremoglu, Ioannis Anagnostakis, Ryan
Rifkin, John-Paul Clarke, John Jensen
NEXTOR
Questions Considered
Has CDM led to improvements in the
quality of information and information
distribution?
What has been the direct impact on GDP
planning at San Francisco and Newark?
Has CDM had an impact on overall airline
decision making?
What are the prospects for future CDM
benefits?
NEXTOR
Inputs to Analysis
Direct analysis of air traffic data
Reports from airlines
Interviews with ATCSCC specialists
NEXTOR
CDM Vision
Improve Air Traffic Flow Management by:
generating better information by combining information
generated by the FAA with information generated by
National Airspace System (NAS) users;
distributing the same information both to FAA managers
and to NAS users;
creating tools and procedures that allow NAS users:
to directly respond to capacity/demand imbalances,
to collaborate with FAA traffic flow managers in the
formulation of flow management actions.
NEXTOR
CDM Status
Agreement on new paradigm for ground delay
programs (GDPs)
Regular meetings of all CDM players
Flight Schedule Monitor (FSM)
CDMNet
Prototype implementation
Work on future applications: NAS Status,
Collaborative Routing
NEXTOR
Questions Considered
Has CDM led to improvements in the
quality of information and information
distribution?
What has been the direct impact on GDP
planning at San Francisco and Newark?
Has CDM had an impact on overall airline
decision making?
What are the prospects for future CDM
benefits?
NEXTOR
The Promise of CDM:
Improved Information Quality
FAA
Airlines
NEXTOR
The Promise of CDM: Better
Information Distribution
Airlines
FAA
Other NAS Users
NEXTOR
Information Quality
Comparison
ETMS: current database of flight information and
predictions (old system) -- monitored via “Cstring”.
ATMS: CDM enhanced database of flight
information and predictions (new system) -monitored via “CDM-string”.
NEXTOR
Predicting Arrival Demand
Accurate prediction of the arrival demand
profile at an airport is essential to the
calibration of a GDP.
departure
time
enroute
time
arrival
at
airport
cancellation
NEXTOR
Information Analyses
Departure Time Predictions
Cancellation Notices
Arrival Time Predictions
Arrival Demand Profile
Under CDM
departure time predictions are based on FAA and airline
provided information
a new cancellation notice is available to the airlines
NEXTOR
IPE metric
Measures performance of a stream of predictions for
a single event
Assigns a single value to each flight over its entire
history
Robust w.r.t. bad flight records - more general than a
snapshot
Can be applied to any stream of predictions for a
single event (dep, arrv, cnx, etc.)
Allows for aggregate stats (e.g. by airline)
NEXTOR
IPE Metric for Departure Time
IPE()=
f
t =n
t=0
ETD(t) =
n
f
k=0
(t
k+1
−tk )perrt
( )
perr(t) = |ETD
A0
A1
t0
t1
t2
t3
NEXTOR
IPE-6 performance (ETD)
Common flights only
Feb2 - Mar16 1998
GDP
Av IPE-6
Av Improvement
CDM
C
CDM
C
SFO
27.21
30.86
13.27
14.44
EWR
28.19
35.40
18.71
6.21
% of Flights Improved
CDM
Equal
C
44.88
39.23
15.89
46.69
31.63
21.69
non-GDP
% of Flights Improved
CDM
Equal
C
23.08
68.00
8.92
23.56
67.49
8.96
Av IPE-6
Av Improvement
CDM
C
CDM
C
SFO
9.74
12.51
9.85
4.21
EWR
13.34
14.72
10.99
12.13
NEXTOR
IPE-6 Improvements at SFO
Feb 2 - Mar 16 1998
C
9%
non-GDP da
C
16%
CDM
23%
GDP day
CDM
45%
Equal
68%
Equal
39%
NEXTOR
0
NEXTOR
* 16-Mar
15-Mar
13-Mar
11-Mar
9-Mar
* 5-Mar
3-Mar
28-Feb
* 27-Feb
26-Feb
25-Feb
24-Feb
* 23-Feb
22-Feb
* 21-Feb
20-Feb
* 18-Feb
* 17-Feb
60
* 16-Feb
15-Feb
* 14-Feb
13-Feb
* 12-Feb
11-Feb
* 10-Feb
9-Feb
* 8-Feb
* 7-Feb
* 6-Feb
* 5-Feb
* 4-Feb
* 3-Feb
* 2-Feb
Percentage of common flights
Strict improvements in IPE-6 performance (on ETD) at SFO
70
CDM better (av = 38.10)
C better (av = 14.15)
50
40
30
20
10
Cancellation Notices
Airlines can cancel flights for a variety of reasons.
The number of cancellations varies substantially
from day to day and can be particularly high in the
presence of GDPs.
During the Jan -- May 1998 period, on 79% of the
days at SFO and 62% of the days at EWR, there
was at least one period of heavy cancellations (4
hour period with 28 or more cancellations).
NEXTOR
-30
NEXTOR
sfo5-31-98
sfo5-26-98
sfo5-21-98
sfo5-16-98
sfo5-11-98
sfo5-06-98
sfo5-01-98
sfo4-26-98
sfo4-21-98
sfo4-16-98
sfo4-11-98
sfo4-06-98
sfo4-01-98
sfo3-27-98
sfo3-22-98
sfo3-17-98
sfo3-12-98
sfo3-07-98
sfo3-02-98
sfo2-25-98
sfo2-20-98
sfo2-15-98
sfo2-10-98
sfo2-05-98
sfo1-31-98
sfo1-26-98
sfo1-21-98
sfo1-16-98
sfo1-11-98
sfo1-06-98
sfo1-01-98
Number of CNX
Daily CNX Volatility
Daily CNX totals
SFO
220
170
120
70
20
0
Hour
NEXTOR
3200
3100
3000
2900
2800
2700
2600
2500
2400
2300
2200
2100
2000
1900
1800
1700
1600
1500
1400
1300
1200
1100
1000
900
800
Number of CNX
CNX Volatility over a Day
CNX by the hour
Sample Day: 1-6-98
25
20
EWR
SFO
15
10
5
SFO Cancellation Notices
Comparison between CNX Notices with
or w/out CDM (TO default: 120 mins after OETD)
3000
2500
w/o CDM (ave = -49)
CDM (ave = 44)
1500
1000
500
0
66
0
57
0
48
0
39
0
30
21
0
0
12
30
0
-6
50
-1
40
-2
30
-3
20
-4
10
0
-5
Frequency
2000
Bin
NEXTOR
Cancellation Notice Summary
Statistics for Jan -- May Period
Under CDM flight cancellations were
reported an average of 47 min before ETD
at EWR and an average of 44 minutes
before ETD at SFO.
Without CDM, we estimate that flight
cancellations would have been reported, on
the average, between 29 and 64 min after
ETD at EWR and 19 and 49 min after ETD
at SFO.
NEXTOR
Significance of Cancellation
Notices
Advance notice of cancellations is particularly
important for GDP planning.
Improved cancellation information was cited by one
ATCSCC specialist as the biggest benefit of
CDM.
Based on the data analyzed to date, the most
dramatic improvement in information quality due
to CDM, is in the area of cancellation notices.
NEXTOR
Questions Considered
Has CDM led to improvements in the
quality of information and information
distribution?
What has been the direct impact on GDP
planning at San Francisco and Newark?
Has CDM had an impact on overall airline
decision making?
What are the prospects for future CDM
benefits?
NEXTOR
Measurement of Impact on
GDP Planning
Measurement of CDM’s overall impact on GDP planning
is difficult because:
CDM influences decision-making in many, sometimes subtle
ways
measuring overall efficiency of a GDP is difficult
Compression algorithm
unique to CDM
eliminates vacant slots and improves overall efficiency
compression impact on assigned ground delay can be
quantified
NEXTOR
Delay Reduction Due To
Compression
During the period of January through May, the use of
compression resulted in an average reduction in
assigned ground delay of approximately 13% at SFO
and 12% at EWR.
Slightly over half of this reduction could have been
obtained by the airlines through substitutions.
The remainder (6% SFO and 5% EWR) could only be
obtained using compression.
NEXTOR
1/
23
/1
99
1/
8
23
1
/1
SF
99
O
8
1/
2
SF 26/1
99
O
1/
8
27
/1
SF
9
98
O
2/
SF 1/19
98
O
SF 2/2/
19
O
98
2/
SF 3/19
98
O
2/
1
3/
SF 199
8
O
2
SF 2/4/
1
O
99
2/
8
SF 5/19
98
O
2/
1
5/
19
SF
98
O
2
2/
SF 6/19
98
O
2/
SF 7/19
98
O
2/
SF
8/
O
19
2/
98
SF 10/1
99
O
2/
8
10
1
/1
SF
99
O
8
2/
2
SF
12
/1
O
9
3/
98
16
/1
99
8
1
O
O
SF
SF
Percentage
Compression SFO
25.00
20.00
15.00
%Red Cmp
%Net Red Cmp
10.00
5.00
0.00
NEXTOR
Value of Compression Savings
Using an industry-accepted value of $25 per
minute of delay, the compression savings
was $26,000 per GDP at San Francisco
and was $29,000 per GDP at Newark. The
respective average monthly savings were
$269,000 and $100,000.
NEXTOR
Airborne Delays
Within a GDP there can be a delicate balance
between assigned ground delay, airborne
delays and throughput.
Issues:
What has been the impact of CDM on airborne
delays?
Do some of the ground delay savings represent a
transfer of ground delay to airborne delay?
NEXTOR
Airborne Delays:
Average Delay per Flight for SFO
all
GDPnon-G
Jan -- March 97
-0.4472.068-0.8
Jan -- March 98
2.111 7.876 0.26
Nov 97
1.647 6.234 1.16
NEXTOR
Airborne Delay: Conclusions
There have been substantial increases in
airborne delay from Jan - March 1997 to Jan
- March 1998.
Majority of increase was already evident in
Nov 1997 (pre - CDM).
We are unable to conclude whether CDM has
had a negative or positive impact on
airborne delay --- more analysis is required.
NEXTOR
Delay Cost Reduction
One could argue that the GDP improvements provided by
CDM are more geared toward allowing the airlines to
reduce the cost of delays rather than only reducing the total
delay within a single GDP. We have not been able to
estimate such savings in this initial analysis.
Continuing flight
45 min delay
15 min delay
Terminating flight
NEXTOR
Evidence of Airline Delay Cost
Savings
United Airlines reports that it has achieved
significant delay cost reduction based on
the use of GDP-E at SFO and EWR and
also the use of FSM to plan its responses to
GDPs at ORD. They estimate the value of
the total savings over the initial 1 1/2
months of prototype operations to be
between $3 to $4 M.
NEXTOR
Improvements in Overall GDP
Efficiency
The majority of the ATCSCC specialists
interviewed felt that, under CDM, they were
able to produce better GDPs:
FSM revision feature was used very effectively.
Power run feature has enabled better decisions on
which centers to include in programs.
Improved data (demand predictions) has helped
design more effective GDPs.
NEXTOR
Impact of Distribution of CDM
Information to Airlines
GDP planning at “non CDM” airports.
Airline flow management.
Fuel Planning.
NEXTOR
GDP Planning
United Airlines has used information provided by
CDM/FSM to help determine the number of
flights to cancel at ORD under conditions of
degraded capacity.
According to UAL, on at least two separate
occasions, the number of canceled flights was
reduced by 25% over the number that would have
normally been canceled; the estimated total cost
savings was $1.5 M.
NEXTOR
Airline Flow Management
US Airways uses FSM to determine the size
and characteristics of heavy arrival banks at
hubs; this information is used by dispatchers
for planning purposes.
In times of very high demand US Airways has
used FSM to implement its own “internal
GDP” to prevent grid lock at a hub airport.
NEXTOR
Airline Flow Management
Delta uses FSM to estimate anticipated
airborne delays on flights arriving into hubs
during peak periods; this is used to
determine whether diversions will be
necessary.
Delta estimates that, based on the more
accurate information provided by FSM, they
have been able to allow flights that normally
would have been diverted to go on to their
destination airports.
NEXTOR
Fuel Planning
United Airlines, US Airways and TWA report
using FSM to estimate airborne delay and
thus obtain a more accurate estimate of fuel
requirements.
NEXTOR
Questions Considered
Has CDM led to improvements in the
quality of information and information
distribution?
What has been the direct impact on GDP
planning at San Francisco and Newark?
Has CDM had an impact on overall airline
decision making?
What are the prospects for future CDM
benefits?
NEXTOR
Prospects for Future Benefits
We feel the following factors indicate the prospect
for future benefits is strong:
movement out of prototype operations
the current extension of CDM-based GDPs to all major
US airports
improved ability of airlines to take advantage of CDM
capabilities
improvements in GDP planning through better data
quality and new FSM features
application of CDM in other areas, including distribution
of NAS status information and collaborative routing.
NEXTOR