Generating Diverse Reroutes for Tactical Constraint Avoidanceatmseminar.org/...ATM_RD_Seminar_presentation_104.pdfSheng Liu . Timothy Stewart . Craig Wanke . The MITRE Corporation

© 2017 The MITRE Corporation. All rights reserved. Add disclaimer here if needed or delete

Christine Taylor Sheng Liu Timothy Stewart Craig Wanke The MITRE Corporation McLean VA

Generating Diverse Reroutes for Tactical Constraint Avoidance

Center for Advanced Aviation System Development

12th USA/Europe ATM R&D Seminar

Seattle, WA 25-29 June 2017

| 2 |

© 2017 The MITRE Corporation. All rights reserved. Add Disclaimer Here if needed or delete

Automation support for tactical rerouting

The Tactical Rerouting Problem

Develops trajectories which factor in aircraft equipage, NAS performance objectives, and operator preferences

2

Provides an alert when flight(s) are affected by weather, sector congestion, SAA(s), equipment outages or delay saving opportunity

1

3 Facilitates collaboration to refine and approve trajectory modifications

Disseminates approved trajectory modifications leveraging XFS/ABRR and Data Comm

4

!

4 descend to FL240

3 climb to FL350

1

2

| 3 |


Challenges for Algorithmic Development

!

Problem Set [set of flights needing

reroutes]

Advisory Set [constraint-avoiding

reroutes for each flight] Advisory Set [constraint-avoiding


reroutes for each flight]

Weather Avoidance

Airspace Conformance

Congestion Reduction

Maintain Schedules

Limited Coordination

Capacity Utilization

Acceptability Metrics

| 4 |


Approach

Problem Set [set of flights needing

reroutes]

Advisory Set [constraint-avoiding



reroutes for each flight]

Generate Candidate Reroutes

Construct Networks

Optimize Reroutes

Flight-specific reroutes


Optimize Advisory Sets Evaluate


Do until Termination criteria is reached

Select Advisory Set

reroutes

Identify Representative Advisory Sets Identify Principal

Components

Cluster and Select

Representative

| 5 |


Generate Candidate Reroutes

Generate flight-specific reroutes using network optimization – Network constructed from fix-pair

segments derived from historical routes

– Arc costs capture desirable behavior – Leverage efficiency of Dijkstra’s

shortest path to generate reroutes Multiple candidates produced using

DSP variant – DSP-M produces the shortest path

through a specified intermediate node “M”

– Forces diversity in the reroutes – Encourages “flows” in multi-flight

examples

!

| 6 |


Metrics of Acceptability

Individual Flight Metrics Design Acceptability

– Distance – Flow conformance

Management Acceptability – Coordination – Return to original route – Number of segments

Constraint Avoidance – Route blockage – Blockage probability – Congestion

Flight Operator Acceptability – Schedule integrity

Advisory Set Metric Flights in Flow

– Computes the maximum number of consecutive fixes common between 2 reroutes

– Longest Common Substring (LCS), where each fix is considered a character

Flight O ABC DEF GHI JKL MNO

Flight B STU DEF GHI JKL

Flight O STU PQR GHI JKL MNO

Flight B STU PQR JKL

ABC DEF GHI

JKL

MNO PQR

STU

| 7 |


Optimize Advisory Sets

Genetic Algorithms (GAs) are a class of heuristic search methods – Mimic biological evolution to identify

good solutions – No guarantee of optimality

Multi-objective GAs (MOGAs) generate the Pareto front of solutions – Each solution defines an Advisory Set – Returns all non-dominated Advisory Sets

Flight A Flight B Flight C

Advisory Set chromosome

Metrics Flight A

Advisory Set Metric

Metrics Flight B Metrics Flight C Total Flight

Metrics

| 8 |


Identify Representative Advisory Sets

Source: http://www.joyofdata.de/public/pca-3d/

MOGA generates a large number of solutions – High-dimensional (10-D) Pareto Front – Clustering in 10-D can obscure key

trade-offs Cluster using Principal Components

Analysis – PCA identifies the primary

correlations between metrics

| 9 |


Identify Representative Advisory Sets

MOGA generates a large number of solutions – High-dimensional (10-D) Pareto Front – Clustering in 10-D can obscure key

trade-offs Cluster using Principal Components

Analysis – PCA identifies the primary

correlations between metrics Select representative from each PCA-

derived cluster – Compute the “central design” – Select solution with minimum

“distance” to central design

2

1

PCA 1

PCA

2

3

A1

B1

C1

A1

B2

C1

A1

B3

C2

A1

B1

C1

Central Design

A1

B3

C1

3/3

1/3

2/3

Freq.

Dist1 = 1+0+2/3

Dist2 = 1+0+2/3 Dist3 = 1+1/3+0

| 10 |


Multi-flight Example

Flight Origin Destination # Color

A Westchester County (HPN)

Raleigh Durham (RDU) 64 Orange

Red

B LaGuardia (LGA) Raleigh Durham (RDU) 62 Olive

Green

C Reagan National (DCA) Charleston (CHS) 17 Royal Blue

D Norfolk (ORF) Orlando (MCO) 84 Dark Blue

E Raleigh Durham (RDU)

Reagan National (DCA) 90 Coral Pink

F Charlotte Douglas (CLT) LaGuardia (LGA) 275 Lime

Green

G Atlanta (ATL) LaGuardia (LGA) 272 Gray

H Baltimore

Washington (BWI)

Norfolk (ORF) 15 Magenta

I Atlanta (ATL) LaGuardia (LGA) 322 Maroon

MOGA generates 5000 unique Pareto-optimal Advisory Sets

[out of 1.8 *1017 possible options]

| 11 |


PCA Analysis

Design Acceptability

Management Acceptabilit

Flights in Flow

Flight Operator Acceptability

Constraint Avoidance

| 12 |


Improved metrics: Distance, Flow Conformance,

Coordination, Return to Route, Sector Congestion, Blockage Probability & # of Segments

Improved metric: Flights in Flow

Impr

oved

met

ric:

Sche

dule

Disr

uptio

n Im

prov

ed m

etric

: Ro

ute

Bloc

kage

Bett

er

solu

tions

Bett

er

solu

tions

Better solutions

Better solutions

Pareto Clustering

| 13 |


Improved metrics: Distance, Flow Conformance,

Coordination, Return to Route, Sector Congestion, Blockage Probability & # of Segments

Improved metric: Flights in Flow

Impr

oved

met

ric:

Sche

dule

Disr

uptio

n Im

prov

ed m

etric

: Ro

ute

Bloc

kage

Bett

er

solu

tions

Bett

er

solu

tions

Better solutions

Better solutions

Pareto Clustering

| 14 |


Legend

PCA reps.

10-D reps.

Legend

PCA reps.

10-D reps.

Advisory Set 1

Advisory Set 2 Advisory

Set 3

Comparison of Clustering Approaches

| 15 |


Advisory Set Reroutes

Advisory Set 1 Advisory Set 2

Advisory Set 3

| 16 |


Contributions

The goal of this project is to reliably generate diverse, operationally acceptable route alternatives to traffic managers The multi-flight context results in coordinated solutions The algorithm generates flows as viable, not enforced Multiple Advisory sets provide distinct choices to traffic managers

For high-dimensional multi-objective optimization, PCA provides a critical analysis of data correlation – Highlights key trade-offs, otherwise lost in the volume of data

| 17 |


This is the copyright work of The MITRE Corporation and was produced for the U.S. Government under Contract Number DTFAWA-10-C-00080 and is subject to Federal Aviation Administration Acquisition Management System Clause 3.5-13, Rights in Data-General, Alt. III and Alt. IV (Oct. 1996). No other use other than that granted to the U.S. Government, or to those acting on behalf of the U.S. Government, under that Clause is authorized without the express written permission of The MITRE Corporation. For further information, please contact The MITRE Corporation, Contract Office, 7515 Colshire Drive, McLean, VA 22102, (703) 983-6000. The contents of this material reflect the views of the author and/or the Director of the Center for Advanced Aviation System Development, and do not necessarily reflect the views of the Federal Aviation Administration (FAA) or Department of Transportation (DOT). Neither the FAA nor the DOT makes any warranty or guarantee, or promise, expressed or implied, concerning the content or accuracy of the views expressed herein. 2017 The MITRE Corporation. The Government retains a nonexclusive, royalty-free right to publish or reproduce this document, or to allow others to do so, for “Government Purposes Only.”

For Release to all FAA personnel. This document was prepared for authorized distribution only.

It has not been approved for public release.

Fiscal Year:

Outcome Number:

PBWP Reference:

Documents

Generating Diverse Reroutes for Tactical Constraint Avoidanceatmseminar.org/...ATM_RD_Seminar_presentation_104.pdfSheng Liu . Timothy Stewart . Craig Wanke . The MITRE Corporation