1 Transportation Systems Research at University of Maryland http://tep.umd.edu Lei Zhang, Ph.D. Associate Professor Director, National Center for Strategic Transportation Policies, Investments, and Decisions Director, Transportation Engineering Program Department of Civil and Environmental Engineering University of Maryland, College Park Phone: 301-405-2881 Email: [email protected]Agent-Based Methods for Transportation Network Optimization DOE ARPA-E Workshop in San Francisco, CA 03/10/2014 1
21
Embed
Agent-Based Methods for Transportation Network Optimization · Design of Experiments (DoE) Transportation Network Simulation Optimization based on surrogates Model Validation. One-stage
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1Transportation Systems Research at University of Maryland http://tep.umd.edu
Lei Zhang, Ph.D.
Associate ProfessorDirector, National Center for Strategic Transportation Policies,
Investments, and DecisionsDirector, Transportation Engineering Program
Department of Civil and Environmental Engineering University of Maryland, College Park
Vehicle Ownership Household Mid- to long-term Household attributes
Location Choice Household Long-term Household attributes, land use
2
3Transportation Systems Research at University of Maryland http://tep.umd.edu
Theory and Methodology
Traditional: Rational Behavior TheoryWhat agents SHOULD doPerfect information and rationalityOptimizing behavior Maximizing utility, profit, welfare, etc.
e.g. mean travel time minimization using optimal toll rates with box contraints
Simulation outputs
e.g. R-square, RMSE, NRMSE, NMAE, EGO
SimulationBasedOptimization
Jointly optimize multiple operations and planning strategiesUse simulation models for evaluation and now for optimization tooMultiple modes can also be jointly optimized with multiple objectives
Simulation-Based Optimization
6
7Transportation Systems Research at University of Maryland http://tep.umd.edu
Active Corridor Traffic Management
1
2
3
4
DMS
Incident Scenario
7
8Transportation Systems Research at University of Maryland http://tep.umd.edu
Congestion: Baseline Scenario
Exit 29
Exit 30
Exit 31
Exit 32
DMS1
DMS2
DMS3
DMS4
8
9Transportation Systems Research at University of Maryland http://tep.umd.edu
Accident without ATM
!Exit 29
Exit 30
Exit 31
Exit 32
DMS1
DMS2
DMS3
DMS4
!
9
10Transportation Systems Research at University of Maryland http://tep.umd.edu
Accident with ATM
!Exit 29
Exit 30
Exit 31
Exit 32
DMS1
DMS2
DMS3
DMS4
!
DMS
10
11Transportation Systems Research at University of Maryland http://tep.umd.edu
Dynamic Pricing Optimization
11
12Transportation Systems Research at University of Maryland http://tep.umd.edu
Multi-Objective Optimization Results
Average Travel Time Total Toll Revenue
12
13Transportation Systems Research at University of Maryland http://tep.umd.edu
China-Singapore Eco-City in TianjinMultimodal Transportation
Planning and OptimizationTarget year 2020, area 30 km2
Projected 350,000 residentsGreen transportation planning145 TAZs, 556 nodes, 1,770 links9 bus lines and 3 LRT lines 7 population groups, 7 activity pairs and 5 travel modes (Bus, rail, car, bike, walk)Transportation Planning goal: Public transportation and non-motorized modes > 90% mode share by 2020
13
14Transportation Systems Research at University of Maryland http://tep.umd.edu
Multimodal System Optimization
Optimal strategyBase Case
Optimal [Parking restriction + Car sharing incentive + + Transit fare] for maximum user benefits
Level of Service
14
15Transportation Systems Research at University of Maryland http://tep.umd.edu
Key Challenges: Behavior Data
No useLow use
High Use
15
16Transportation Systems Research at University of Maryland http://tep.umd.edu
Level-of-Service Comparison
Freeways Freeways + ArterialsAverage
Difference11%
(24 stations)15%
(62 stations)
Traffic Count Comparison
Travel Time ComparisonAM Peak PM Peak
Travel Time Difference |∆|
14% (9 corridors)
12 % (9 corridors)
Model Calibration and Validation
16
17Transportation Systems Research at University of Maryland http://tep.umd.edu
Color Ramp:Attribute:
2010 2030 Summary
SHA Agent-Based Model Web Reporting System
Agency and User Support
select intersectionselect one linkselect one superlinkselect multiple linksselect areaselect all
48.753.3
42.245.4
52
35.1
0
10
20
30
40
50
60
Corridor 1 Corridor 2 Corridor 3
Trav
el T
imes
(m
in)
Before ICCAfter ICC
17
18Transportation Systems Research at University of Maryland http://tep.umd.edu
Real-Time Decision Support
Decision-Maker
18
19Transportation Systems Research at University of Maryland http://tep.umd.edu
Development SiteBoston
Application SiteBaltimore
You
DMS
Normal route
Diverting route
Dynamic msg. sign
Bluetooth detector
DMS
Example: En-Route Diversion Model Transfer
Model Transferability
19
20Transportation Systems Research at University of Maryland http://tep.umd.edu
Closing RemarksSimilarity between Energy and Transportation Grids:Agents, Networks, Critical Infrastructure, …
Opportunity: Nonlinear and complex relationships between agent behavior and system performanceSystematic identification of feasible behavior shifts that can produce significant system benefitsModel development should be driven by data availability and analysis needsBig, exciting, but still imperfect dataDecision-makers want more information, better information, and they want it now, in real time
20
21Transportation Systems Research at University of Maryland http://tep.umd.edu
Questions, Comments, and Suggestions are Welcome. Please Contact:
Lei Zhang, Ph.D., Associate ProfessorDirector, National Transportation CenterDirector, Transportation Engineering ProgramDepartment of Civil and Environmental Engineering1173 Glenn Martin Hall, University of MarylandCollege Park, MD 20742Email: [email protected]: 301-405-2881Web: http://www.lei.umd.edu