1. 11 Quantifying Benefits of Traffic Information Provision under Stochastic Demand and Capacity Conditions: A Multi-day Traffic Equilibrium Approach Mingxin Li Xuesong Zhou Nagui M. Rouphail Presentation for the 14th International IEEE Annual Conference on Intelligent Transportation Systems October 7, 2011

2. 2 Motivation Revised from FHWA congestion report http://www.ops.fhwa.dot.gov/congestion_report/executive_summary.htm#what_is_congestion Special Event, 5% Poor Signal Timing, 5% Physical Bottlenecks, 40% Bad Weather, 15% Traffic Incidents, 25% Work Zones, 10% 2/42

3. 3 Stochastic Capacity Stochastic capacity can be defined as the maximum sustainable flow rate 3/42

4. 4 Outline Background and motivation Definition of user types Multi-day solution strategies Summary of numerical experiment

5. 5 Background and Motivation Traffic systems can be viewed as stochastic processes with non-deterministic demand and capacity values Given traffic demand flow and queue profiles on a link or a corridor, key issues are:  Can additional high-quality traffic information provision services improve the system-wide average travel time or travel time reliability?  Can the improved network knowledge quality necessarily improve the overall system performance?  How to quantify the benefit and then prioritize various potential congestion mitigation solutions?

6. 6 Definition of User Type (1) TI users Every day, link travel time estimates with a certain level of prediction errors are available for TI users to make route decisions 12112111... TI User Pre-trip information: can affect Route Destination En-route information: can change Route Destination Driving behavior

7. 7 Definition of User Type (2) TI User With non-perfect or incomplete information, TI User can switch routes every day. TI User ETT user ETT user selects the same route every day, regardless the actual traffic conditions. User #1 User #2 Map Source: Google Map ©2011  No pre-trip traveler information  No en-route traveler information  Based on their knowledge on the average traffic conditions

8. 8 Simple Network Used as an Illustrative Example

9. 9 Day-dependent Capacity, Demand and Travel Time Patterns Day-dependent travel time patterns with 100% ETT users  Full capacity= 4,500 veh/h (80%)  Reduced capacity= 3,000 veh/h (20%)  High demand =9,600 veh/h (20%)  Low demand =7,600 veh/h (80%)

10. 10 Overall Model Formulation  Traveler Information (TI) users  Expected travel time (ETT) knowledge-based solution  Objective function Perception Error Scaling Parameter Path Utility Measured Travel Time

11. 11 Solution Algorithm Step 1: Initialization Generate demand vector & road capacity vector Step 2: Multi-day traffic simulation with stochastic capacity Generate day-dependent link travel times according to stochastic capacity vector Step 3: Find descent directions for stochastic traffic assignment Find the Least Travel time Path (LTP) using day-dependent link travel time on each day d. Find the Least Expected Travel Time Path (LETP) using average link travel time Step 4: Path assignment for TI and ETT vehiclesStep 5: Link flow aggregation For each day, calculate the aggregated link volume using TI flow volume and ETT flow Step 6: Convergence checking Calculate the gap function: If convergence is attained, stop. Otherwise, go to Step 2.

12. 12 Sample Solution for 5% TI Users & 95% ETT Users Legend RC: Reduced Capacity FC: Full Capacity HD: High Demand LD: Low Demand Default perception error scale for ETT users Information accuracy scale for TI users TI market penetration rate

13. 13 Travel Flow Split Solution on Four Different Types of Days

14. 14 System Demand (1) Alameda I-80 - Primary Direction

15. 15 System Demand (2) Demand flow rate during peak-hour

16. 16 MOE at Different Market Penetration Rates

17. 17 Representative Traveler Information Provision & Traffic Management Strategies

18. 18 Contributions Evaluate benefits of traveler information provision strategies in a realistic environment with stochastic traffic demand, stochastic road capacity Consider various sources of travel time uncertainties: o Variations in link capacities and travel demands o Imperfect parameter estimation for link travel time functions o Route choice behaviors Extend the multi-day analysis framework in real-world ATIS applications: o Both demand and supply are stochastic; o Travelers have different degrees of knowledge and information quality across different days

19. 19 Conclusions Evaluate how travelers with different information accessibility adjust their route choice patterns when various sources of travel time uncertainty The proposed model can be also enhanced to consider more realistic route choice utility functions involving both expected travel time and travel time reliability.