Information-based adaptive routing: Path v.s Policy

Transcription

1 Information-based adaptive routing: Path v.s Policy Nam Hong Hoang Supervised by: Prof. Hai Vu & Dr. Manoj Panda Intelligent Transport Systems Lab (ITSL) Centre for Advanced Internet Architectures (CAIA) Swinburne University of Technology Information in ITS etsi.org CAIA Seminar November 19, 215 3

2 The actors Travelers Request information: best routes... Form information: as a part of traffic Generate new information: make incidents... Source of information (V2V) Information providers Process (uncertain) information of traffic states Provide consistent information, used and adjusted by travelers Network operators CAIA Seminar hhoang@swin.edu.au November 19, The dual problems (for operators) Given network and demand, find the optimal strategies to manage traffic Given traffic characteristics and demand, design the optimal network (settings, topology) Other problem: Capacity design Maximize demand given traffic characteristics and network infrastructures Multiple dimensions (time and space): static and dynamic, deterministic and stochastic Adaptive routing with information to reduce uncertainty Require an analysis or simulation method to find solutions CAIA Seminar hhoang@swin.edu.au November 19, 215 5

3 The existing DTA framework DTA Traffic (TM) Choice (CM) DTA: Dynamic Traffic Assignment In general, DTA models are non-linear Non-holding-back, FIFO Solution methods: Heuristics, Fixed-point algorithms, etc. CAIA Seminar November 19, Traffic model: Cell Transmission [1] A link is divided into segments or cells Dynamic description of road segments, caused by incidents Spatial distribution of traffic within each cells is averaged flow Max flow Q density CAIA Seminar hhoang@swin.edu.au November 19, 215 8

4 The proposed analysis framework DTA with Information framework Traffic (TM) Information (IM) Choice (CM) SO- DTA TTM CTM Perfect + Complete Policy choice Path choice Novel contribution: Information model LINEAR approach to the whole framework CAIA Seminar hhoang@swin.edu.au November 19, Policy choice v.s path choice Policy choice: Choosing a next link or cell to move on Temporal-spatial adaptation Path choice: Choosing a path to move on Temporal adaptation? Policy choice??? Path choice CAIA Seminar hhoang@swin.edu.au November 19, 215 1

5 The model settings Traffic model: Cell Transmission Information model: perfect (no error/noise) and complete Routing: policy and path choice Optimization model: Objective: Constraints: Minimize the total travel time CTM constraints Path/Policy choice constraints CAIA Seminar November 19, An example Demand: 48 veh (R to S), all starting at time R 4 S Time unit: 3 seconds Scenarios for cells 2 and 5 Max flow Time period s 16 veh/time unit ALL s 1 8 veh/time unit 8 13 s 2 8 veh/time unit 8 17 Travelers are able to acknowledge s after time 8. Travelers are able to acknowledge s 1 or s 2 after time 14. CAIA Seminar hhoang@swin.edu.au November 19,

6 Scenario s : R S path Path: R S path 1 Policy: CAIA Seminar hhoang@swin.edu.au November 19, Scenario s 1, s 2 : path path Path: Policy: path path CAIA Seminar hhoang@swin.edu.au November 19,

7 Computational performance Nguyen-Dupuis network [2] R1 1 R m 6566 m 1315 m 2816 m 4683 m 47 m Bottleneck m 122 m 9382 m 5632 m Bottleneck 7515 m S m S2 CAIA Seminar hhoang@swin.edu.au November 19, Computational performance Number of constraints Number of variables PoSCTM O((C + A)X TC S ) O((C + A)X TC S ) PaSCTM O((C + A)X TP) O((C + A)X TP) Path Policy = P C S = Number of paths Number of destinations CAIA Seminar hhoang@swin.edu.au November 19,

8 Computational performance Complexity (constraints, variables) PoCTM-1scen-constraints PoCTM-1scen-vars PoSCTM-2scen-constraints PoSCTM-2scen-vars PaCTM-1scen-constraints PaCTM-1scen-vars PaSCTM-2scen-constraints PaSCTM-2scen-vars Time unit (s) CAIA Seminar hhoang@swin.edu.au November 19, Computational performance Execution time execution time (s) PoCTM-1scen-solution PoCTM-1scen-preparation PoSCTM-2scen-solution PoSCTM-2scen-preparation PaCTM-1scen-solution PaCTM-1scen-preparation PaSCTM-2scen-solution PaSCTM-2scen-preparation Time unit (s) CAIA Seminar hhoang@swin.edu.au November 19,

9 Summary Policy-based routing is better than path-based routing Performance Objective value BUT,... Psychological issue: stressful Driver-less car Imperfect and incomplete information What is next? CAIA Seminar November 19, References I [1] C. F. Daganzo. The cell transmission model: A dynamic representation of highway traffic consistent with the hydrodynamic theory. Transportation Research Part B: Methodological, 28(4): , [2] S. Nguyen and C. Dupuis. An efficient method for computing traffic equilibria in networks with asymmetric transportation costs. Transportation Science, 18(2):185 22, CAIA Seminar hhoang@swin.edu.au November 19,