National Airspace System Probabilistic Traffic
Flow Management in Inclement Weather
March 15, 2006
George Hunter, Kris Ramamoorthy, Ben Boisvert,
Ari Stassert, Anna Drabowski, Jim Phillips, Alex
Huang, Fred Wieland, Greg Carr
Sensis Technology Center
1
Probabilistic TFM
Departure
time
Probabilistic TFM
2
Trajectory-based Weather Impact
Analyze specific trajectory, not
just flow through sector
Analyze Wx impact on the specific
trajectory
Reroute
Flight plan
ZKC02 High sector
Count a route as unavailable only if reasonable
reroutes are also unavailable
3
Sector Capacity PDF
Maximum
Sector capacity
Example MSC’ results for
May 10, 2004. Sector
ZKC02. Randomly select 30
flights that transit the
sector and compute upper
and lower bounds. Repeat
this 10 times to check for
convergence.
Aggressive
rerouting
Result: Good convergence
found. 30 flights selected at
random reproduce upper
and lower bound estimates
with low variance.
No rerouting
Time (15 minute steps)
4
Uncertain Storm Forecast
• Result: A forecasted sector capacity
probability distribution function (PDF)
Frequency
– For all sectors, all time horizons, all time slots
0
Capacity
max
5
Computing Airspace Demand and
Capacity Probability
0.4
0.2
0.35
0.3
Probability
Probability
Probability
0.25
0.45
0.25
0.2
0.15
0.1
3
5
Now
7
9
11
0
13
15
17
0
1
19
3
5
0.16
clear
7
9
11
13
15
17
19
1
0.12
0.14
0.1
0.12
0.1
0.08
0.06
0.08
0.06
0.04
Many factors impact PDF shape
0
0
1
3
5
7
9
11
13
15
17
19
1
3
Predicted Sector Capacity
5
7
9
11
13
15
17
1
19
0.07
0.06
0.06
0.06
0.05
0.03
0.05
0.04
0.03
0.02
0.02
0.01
0.01
0
0
3
5
7
9
11
13
15
Predicted Sector Capacity
17
19
Probability
0.07
0.04
9
11
13
15
17
19
5
7
9
11
13
15
17
19
17
19
Predicted Sector Capacity
0.07
1
3
Predicted Sector Capacity
0.08
0.05
7
0.1
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0.08
Probability
• Long look-ahead time
• Medium convective
weather
• Inaccurate storm
forecast
0.02
0.02
Probability
Short look-ahead time
Clear weather
Extreme weather
Accurate storm forecast
5
Predicted Sector Capacity
0.04
•
•
•
•
3
Predicted Sector Capacity
Predicted Sector Capacity
Capacity
0.1
0.05
1
0
0.15
0.05
Probability
Capacity
clear
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Probability
0
Sector Capacity Probability
Probability
Frequency
Frequency
Sector Capacity Probability
0.04
0.03
0.02
0.01
0
1
3
5
7
9
11
13
15
Predicted Sector Capacity
17
19
1
3
5
Hours
from
now
7
9
11
13
15
Predicted Sector Capacity
Capacity uncertainty grows
dramatically with length of look-ahead
6
Example Sector Loading PDF
0.14
60 minute look ahead time
12 aircraft predicted
6 have departed
Probability
0.12
0.1
0.08
0.06
0.04
0.02
0
1
3
5
7
9
11
13
15
17
19
Predicted Sector Loading
7
Congestion Cost Time History
Predicted loading overlaps
predicted capacity causing
non-zero congestion cost
Mean predicted capacities
Mean predicted load
(at gate and en route)
8
Congestion: Load vs Capacity
Predicted
load
Predicted
capacity
Predicted Weather
Congestion
probability
Congestion map: a combination
of the predicted weather impact
and predicted traffic
27 July 2004
9
0145 MidAtlantic, 27Jul04
10
2200 Congestion Map
11
ETMS Reroute Count
Scaled to 47k flights
~1% flights rerouted > 30% path stretch
12
Significantly Fewer Rerouter With
Our Flight Planner
264 reroutes
13
ETMS Congestion Profile
#Overloaded sectors = 1770
14
Congestion/Delay Tradeoff
Comparison
Today’s system
(~1,770 congestion events
and ~25 minutes delay)
#Congestion events
1000
800
600
Predeparture delay
only
400
Predeparture delay
& rerouting
200
Predeparture delay only WRT MAP
0
10
20
30
40
Delay (minutes)
50
60
70
15
ETMS Congestion Distribution
Mean overload = 138%
16
Mean Overload Comparison
Mean congestion (ops)
6
ETMS mean overload count = 5.5
5
4
3
2
1
10
20
30
40
50
60
70
Delay (minutes)
17
An Improved NAS
Today’s technology level
Pr
ob
ab
ili
st
ic
TF
M
so
l
ut
io
ns
Reduced congestion and
delay using trajectory-based
Probabilistic TFM
18
Traffic Planning with Uncertainty
Weather impacts airspace, but the
forecasted capacity and loading are
uncertain.
The result:
Robust traffic
flow in the
future NAS.
Frequency
45
40
35
Perfect
Persistence
30
25
Actual
20
15
10
5
0
Trajectory-based traffic flow
optimization using dynamic
congestion map.
0
Observed traffic
0
1
2
Mean >30 dBZ Dwell Time (min)
Predicted
load
Congestion
probability
3
Predicted
capacity
Capacity
max
Traffic load
exceeding capacity
=> congestion.
19
Backup
20
Probabilistic TFM
• Result: A forecasted sector capacity
probability density function (PDF)
Frequency
– For all sectors, all time horizons, all time slots
0
Capacity
max
21
Mean Overload Comparison
Mean congestion (ops)
6
ETMS mean overload count = 5.5
5
4
3
2
1
10
20
30
40
50
60
70
Delay (minutes)
22
Traffic Planning with Uncertainty
Frequency
Weather impacts sector airspace,
but the forecasted capacity and
loading are uncertain.
45
The result: A
robust traffic
flow in the
future NAS.
40
35
Perfect
Persistence
30
25
Actual
20
15
0
10
5
0
Trajectory-based traffic
flow optimization using
dynamic congestion map.
Observed traffic
0
1
2
Mean >30 dBZ Dwell Time (min)
Predicted
load
Congestion
probability
3
Predicted
capacity
Capacity
max
Traffic load
exceeding
capacity =>
congestion.
23