By using DTA, you accept the following assumptions

Transcription

1 Modeling Express Lanes Using Dynamic Traffic Assignment Models Yi-Chang Chiu, PhD DynusT Laboratory University of Arizona Florida DOT Managed Lane Workshop May, 03 DTA Assumptions By using DTA, you accept the following assumptions Estimating lane demand is a route/departure choice, not a mode choice problem. Credit: Jeff Shelton, TTI

2 DTA Assumptions By using DTA, you accept the following assumptions Learning and adaptation is part of route choice decisions (in lieu of instantaneous or reactive route choice behavior). Day Day Try Learn Anticipate Day n (final) DTA Components for Pricing Analysis 4 Pricing model (optional, depending on pricing scheme) Route/lane choice model Departure time choice model Developed separately Loop back to demand model Network needs to large enough to include most trips through the facility. A small area overpenalizes tolled facility.

3 Sub-Area Scope Implication 5 Why DTA for Express Lane Analysis Given We have had Macro-MicroMicro Macro-Micro inconsistency Traffic dynamic, route choice behavior V/C >.0 V/C <.0 + Queue +.. Learning, static Reactive, stochastic 6 3

4 Why DTA for Express Lane Analysis Given We have had Macro-MicroMicro Better consistency with Macro-Meso-Micro integration V/C >.0 V/C <.0 + Queue +.. V/C <.0 + Queue +.. Learning, static Learning, dynamic Reactive, dynamic 7 Why DTA for Express Lane Analysis Given We have had Macro-MicroMicro Critical Bridge from Macro to Micro and feedback 8 4

5 Pricing Schemes in Real World Samples of a wide range of configurations Facility HOV SOV/Two- Axial HOV Lanes Commercial/3 + Axial (Free) (Very High) (Very High) 9 DUE DUE/ Non-Eq Toll Roads - Fixed rate (Free/Normal/ - ToD rate Low) Managed Lanes - Fixed rate - ToD rate - Congestion Responsive (Normal) (High) (Free) (Varying) (Varying, higher) Analyzing Tolling in a Simple Static Case T(00) = 0 min T(00 $) = 0 min + $/0*60 = 6 min T(50) = 0 min T(50) = 0 min T(80 $) = 7min + $/0*60 = 3 min T(70) = 3 min 0 5

6 Dynamic Case Fundamental Total Experienced Time = 3 min Total Experienced Time = 5 min Static/Instantaneous View of the Same Situation Total Instantaneous Time = min Total Instantaneous Time = 9 min 6

7 Dynamic Case With Static Toll Total Experienced GC = 3. min Total Experienced GC = 5 min 3 Dynamic Case With Dynamic Toll Total Experienced GC = 3.5 min Total Experienced GC = 5 min 4 7

8 Revenue/Throughput Maximizing Pricing Model Pricing Model (vary by software) 5 5 Congestion Responsive Tolling Credit: Jeff Shelton, TTI 8

9 Congestion Responsive Tolling Credit: Jeff Shelton, TTI Maximize throughput while maintaining target speed Denver-Boulder US36 Congestion Pricing TIGER Applications Modeled 3 buildout strategies Estimated Congestion Revenue Travel time Fuel consumption Emission 9

10 Zone: 83 Nodes: 0,095 Links: 3,47 DRCOG Regional Model in DTA AM (6:30 0:30).56M veh PM (3:30 7:30) 3.3M veh 9 US 36. Managed Lane Sub-Area 0

11 Toll Rate per hour ($/hr) $.00 $.80 $.60 $.40 $.0 $.00 $0.80 $0.60 $0.40 $0.0 $00 Million Build (AM) GP Lane Congestion vs. ML Price Federal Sheridan Church Ranch Wadsworth $0.00 Time of Day Federal Sheridan Church Ranch Wadsworth Interlocken No Build $60 Million Build Addressing Heterogeneity Value-of-Time (VoT) plays a critical role in generalized cost type of DTA approach. VoT differs by trip purpose/socio-economic/ location attributes. Increasing interesting in DTA and Activity-Based Model (ABM) integration ti ABM produces trip/person specific VoT.

12 3 Addressing Heterogeneity Existing approaches Discrete Choice Model - Individual route choice - Utility functions for toll and nontolled routes Pros - Flexible, easy to implement - Computational efficient Cons - Hard to converge - Implication/interpretation Multi-Class/Stratified Assignment - Stratify population by VoT distribution - Multi-class assignment for each VoT strata Pros - Seek to converge - Stable results for scenario comparison Cons - Computational demanding and becomes intractable with increasing problem size Summaries - Toll modeling with DTA Include alternate routes to model diversion Static models do this well. Microsim not suitable for equilibrium DTA. Network too small. Not solving time-dependent shortest path. DTA networks need to be large enough. Local congestion can affect tolls from far away Static models do not do queues and spillback. Microsim has lots of detail. DTA combines sufficient realism with largerarea networks.

13 Summary Advantage of applying DTA for pricing analysis appears to be obvious (hopefully). Various approaches are being offered for the following pricing schemes Fixed toll Time-of-day toll Congestion responsive Addressing heterogeneity is desired Ongoing research (SHRP L04, C0, etc.) Importance of feedback 5 3