Lessons Learned from FAA Free Flight's En Route Modeling Workshop with a CPDLC Modeling Case Study

Dave Knorr & Mike BennettFAA Free Flight Office

4 November, 2003

Need for En Route Modeling

• Future demand will begin to increase delays related to en route sector capacity constraints

• Need capability to assess the magnitude of the en route problem

• Need modeling to compare the System level value of alternative solutions (datalink vs. ERAM vs. airspace redesign vs. XXX)

• Need to be able to develop credible $$ benefit projections

En Route Modeling Workshop Overview

• FAA Free Flight Office held a modeling workshop in March of 2003

• Goals were:– understand modeling options for en route sector

capacity increases – Insight into consistency of model results – Assess FACET as potential model engine

• Attendees– Boeing, NASA, CENA, Mitre CAASD, LMI, Metron, SAIC,

CSSI, Aerospace, NEXTOR, FAA Tech Center, etc.

Modeling Workshop Findings

• No single model met all of the needs• List of functionalities of models• No understanding of the relative validity of

results• Need for establishing a standardized set of inputs

for model comparisons

What should a model include? A case study

• From a user’s perspective, what capabilities should a model have?

• What a model needs to do:– Quantify benefits of tool to airspace user

• Delay reduction, fuel savings– Estimate benefits over lifetime of tool (10 years)

• Breakdown by Center, by Year

• Consider Controller-Pilot Data Link Communications (CPDLC) Investment Analysis as a case study

Controller Pilot Data Link Communications (CPDLC)

CPDLC Benefit Analysis Steps

• Determine Workload impact of CPDLC– Performed with voice tape analysis, HITL simulation

• Estimate Sector Capacity Increase (High vs. Low Sectors)• Model decrease in delay associated with

– capacity increase– increased trajectory efficiency

CPDLC Reduced Voice Occupancy

Increased En Route Capacity Increased Voice Channel Availability

More efficient conflict

resolution

Reduced delay in good weather

Reduced delay in severe weather

NASPAC & DPAT CSSI TTT

Voice Channel Utilization with CPDLC

• CPDLC– Enhances En Route Controller productivity by automating

communication which increases sector capacity– Reduced workload improves controllers’ efficiency in vectoring for

traffic

Total Voice Channel Utilization

01020304050607080

0 50 100 150

Fraction of MAP

Per

cent

Voi

ce C

hann

el

Utili

zatio

n

ZDV 25ZDV 38ZDV 3ZDV 14ZDV 30ZTL 6ZTL 4ZTL 3ZFW 28

Voice Channel Utilization for CPDLC Messages

01020304050607080

0 50 100 150

Fraction of MAP

Perc

ent V

oice

Cha

nnel

Ut

iliza

tion

ZDV 25ZDV 38ZDV 3ZDV 14ZDV 30ZTL 6ZTL 4ZTL 3ZFW 28

Model for Sector Capacity Increase

0%

20%

40%

60%

80%

100%

Number of flights in sector

Perc

ent W

orkl

oad

Non-communications tasks

CommunicationsWith CPDLC

No CPDLC With CPDLCMaximum sector capacity

CommunicationsNo CPDLC

Modeled Delay from NASPAC

• Advantages of NASPAC/DPAT Model– Captures sector capacity constraints– Allows full NAS Modeling– Supports Growth in Demand

NASPAC Daily En Route DelayHigh Sectors

01000200030004000500060007000

Baseline 5% 15% 20% 30%Capacity Increase

Ave

rage

Dai

ly

Del

ay (m

in)

200220102020

NASPAC Daily En Route DelayLow Sectors

01000200030004000500060007000

Baseline 5% 15% 20% 30%Capacity Increase

Ave

rage

Dai

ly

Del

ay (m

in) 2002

20102020

Good Weather Using Southern Region Routing

Modeling Problems Encountered

• Issues Within NASPAC/DPAT– TRACON Capacity– Severe Weather

• Issues with Input Assumptions– Routing Assumptions Sensitivity

• Issues not Addressed by NASPAC/DPAT– Efficiency when Demand less than Capacity

Problems encountered: TRACON Capacity

• In reality, airport congestion affects en route airspace

• Assigning finite capacity to TRACON airspace in model allows this “backup” to be simulated

• Tricky to get right– No well-defined standard for what that capacity should

be– Most modelers decouple airport constraints from

enroute constraints

Implementing TRACON capacity

Modeled delay with and without Tracon Capacity

0

5000

10000

15000

20000

2002 2010 2020

Tota

l dai

ly e

nrou

te d

elay

(m

inut

es)

Unlimited29% of AAR

Modeled delay with and without Tracon capacity

0100020003000400050006000700080009000

Baseline 5% 15% 20% 30%Increase in Enroute Sector Capacity

Tota

l Dai

ly E

nrou

te

Dela

y (m

inut

es)

Unlimited29% of AAR

Tracon Capacity, Major Airports March 7, 2002

0%10%20%30%40%50%60%70%

ATL

BO

SB

WI

CLE CLT

CV

GD

CA

DE

ND

FWD

TWE

WR

FLL

IAD

IAH

JFK

LAS

LAX

LGA

MC

OM

DM

IAM

SP

OR

DP

DX

PH

LP

HX

PIT

SA

NS

EA

SFO SLC STL

TPA

Num

ber o

f AC

in T

raco

n (%

of h

ourly

AA

R) max

95th %ile

Average of 95th

percentile:29%

Severe Weather Modeling

• Routing around severe weather increases number of areas where demand exceeds capacity

• Available models did not allow for routing around affected areas

• We used actual tracks as input trajectories– Weather has already been routed around, which may not

have been necessary if more capacity had been available

NASPAC Bad Weather En Route DelayHigh Sectors

0

5000

10000

15000

20000

Baseline 5% 15% 20% 30%Capacity Increase

Bad

Wx

Dai

ly D

elay

(m

in)

200220102020

NASPAC Bad Weather En Route DelayLow Sectors

0

5000

10000

15000

20000

25000

30000

Baseline 5% 15% 20% 30%Capacity Increase

Bad

Wx

Dai

ly D

elay

(m

in)

200220102020

Bad Weather based on tracks “as flown”

Sensitivity to Routing Assumptions

• This is not necessarily a modeling issue per se– Underlines necessity for standardized scenarios

• Observed delay for given capacity in future scenarios is very sensitive to

– Routing assumed• Great circles, pref routes, RNAV routes

– Details of future scheduling• Forecast (TAF) gives ops counts, but not city-pairs• What about rolling hubs? More point-to-point?

Routing Scenario 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017Southern Region routing 0 0 212 445 639 831 987 1333 1679 2042 2418 2852 3462 3901All Great Circle routes 0 0 2546 6697 10254 14678 18291 27840 37307 46835 56629 67232 79364 90873Most likely value 0 0 424 891 1277 1661 1974 2665 3358 4084 4835 5704 6924 7803

YearModeled Daily Delay Savings in 8 CPDLC ARTCCs (minutes per day)

Modeled Sector En Route Daily Delay

2012

Capacity constraints in en route airspace will become more of a

problem in the future

Many en route sectors are currently capacity constrained

High Sectors

2002

Trajectory Efficiency

• Queue-based enroute models apply no efficiency penalty when demand < capacity

• In practice, efficiency of conflict resolution and descent/ascent profiles vary with sector load

• CSSI (P. Lucic, S. Mondoloni, W. Weiss) constructed an ad hoc model built on TTT to estimate benefit

High and Superhigh Sectors

y = 0.0762x + 1.4182

02468

1012

0 10 20 30 40 50

Voice Channel Occupancy (%)

Exce

ss D

ista

nce

(nm

i)

Distance at lower alt vs voice occupancy

y = 15.673x + 1.5912

0

5

10

15

20

25

0 0.2 0.4 0.6 0.8 1

% of time voice channel occupied

dist

ance

trav

eled

(nm

)

En Route Efficiency Pool – ‘Excess’ Distance

Efficiency measure

Here’s what we’d like to see a model do…

Modeling Airspace Performance

If sector load > capacity,

• Allow to enter and adjust dwell time, OR

• Delay

If delay is excessive, adjust trajectory—HOW???

If sector load < capacity,

Adjust dwell time

Excess Distance vs. Sector Load

0

0.5

11.5

2

2.5

3

0 20 40 60 80 100 120 140

Percent Load (1 Min Bins)

Avg.

Exc

ess

Dist

ance

Modeled aircraft approaches new sector

Modeling Airspace Performance

How to do route adjustment to avoid excessive delay?

Dynamic:Route around congested area

Iterative:Limited set of alternate full trajectories

Modeling Airspace Performance

What about terminal delay that can’t be routed around?

In the case of excessive terminal delay, crossing traffic is affected.

This can happen, but not always and is hard to model.

Hold on ground,Then release on same trajectory?

Hold

Backup Slides

What is CPDLC?

• CPDLC allows digital text messaging between enroutecontrollers and pilots

– Replace the voice messages:• Initial Communications• Transfer of Communications• Altimeter Setting

– Also transmits locally-defined menu of common messages• Benefits Include:

– Reduced voice channel occupancy• Greater communications availability More efficient maneuvers• Decreased controller workload Increased sector capacity

– Increased communications accuracy• Reduced readback errors• Reduced OD/OE

Model Needs 1: Input

• A model should compute trajectories of aircraft from input files:– Parse flight plans to produce route of flight– Include wind effects in modeled trajectory– Use individual airplane characteristics to model aircraft

performance– Compute pierce and dwell times for all airspace

elements for each aircraft– Allow linkage of different flight legs

Model Needs 2: Airspace

• A model should simulate the behavior of the full airspace:– Include enroute centers, terminal areas and runways– Include capacity limits on airspace elements,

specifically sector and TRACON flight count limits and runway operation rates for airports

– Include demand-dependent airspace performance based on real-world data (i.e. increase in excess distance in enroute sectors when busy, change in runway service times at different demand levels)

Model Needs 3: Dynamic Adjustments

• A model should allow dynamic adjustments to aircraft trajectories based on the current state of the airspace:– Include the ability to reroute aircraft at different levels of

control• Automatic rerouting around congested sectors, to emulate

behavior of controllers and traffic management units• Strategic rerouting as part of traffic flow management initiatives

– Include ability to implement traffic flow management initiatives, such as ground delay programs

– External perturbations to the airspace, such as convective weather or security issues, must the handled by the model

Model Needs 4: Output

• A model should generate detailed reports, both summary (full-run) and at selected sampled times:

• Sector and TRACON loads• Aircraft trajectories• Aircraft delay