Practical Use of ADUS for Real- Time Routing and Travel Time Prediction

Transcription

1 Practical Use of ADUS for Real- Time Routing and Travel Time Prediction Dr. Jaimyoung Kwon Statistics, Cal State East Bay, Hayward, CA, USA Dr. Karl Petty, Bill Morris, Eric Shieh Berkeley Transportation Systems, Inc., Berkeley, CA, USA November 2005

2 Motivation Trip from Berkeley to Stanford

3 Overview Archived Data User Service (ADUS) system PeMS (Freeway Performance Measurement System) A prototype dynamic route guidance system robust real-time and historical traffic content from PeMS path finding and route guidance algorithm Advanced Traveler Information Systems (ATIS). Benefit of the real-time route guidance system For a simplified freeway network in Los Angeles, which shows an average of 30~60% reduction in travel time for trips that last an hour or longer.

4 ADUS System and PeMS ADUS system Archives ITS-generated operational data for such uses as planning, performance monitoring, program assessment, policy evaluation, safety, and research. Data stored in ADUS are typically characterized by Huge quantity Wide geographical coverage Wide temporal coverage PeMS ( ) The current state-of-the-art example of ADUS systems On each day, PeMS archives 2 GB of data collected and processed from over 26,000 detectors in California in real-time. It covers most urban freeways in California and spans about 5 years in duration.

5 Route guidance system Given a query consisting of the origin and destination (O-D), a route guidance system provides turn-by-turn directions as well as the duration of the trip for a traveler making the trip between the O-D Find (near-) optimal routes using a variant of shortest path algorithms applied to the road network database Such service is typically provided as a function of a geographic positioning system (GPS) unit or via web-based system such as Mapquest (2). Other Systems Route Travel Tim e System Interface: VAR real-time data feed, CMS travel time feed Freeway Database VARS Raw Data Processed Data Speeds AAA Database Transportation System Database Data Grinding Layer: Diagnostics, Filtering, Im putation, Aggregation Incident Database Transformation Logic Weather Database Web Users Planners Engineers Administration Academic Public Web Site: User Applications, Presentation Logic, Adm inis trative Control Users Interface Layer Data Warehouse Layer Data Staging Layer Freeway Data Sources Incident Data Sources Weather Data Sources Data Sources

6 Dynamic route guidance system Currently, most route guidance systems use the nominal travel time based on the speed limit for path finding as well as for trip duration computation. useful, but not satisfactory Little practical effort in implementing such a system that incorporates real-time traffic information in shortest path Primarily due to the lack of available data Our system is one of the first robust and mature webbased dynamic route guidance systems

7 System components PeMS Coding freeway network in PeMS into segments (~1/2 mile). Historical average travel time that is specific to the time of day and day of week. Real-time travel time information Shortest path algorithm The road network represented as a weighted directed graph (di-graph) Dijkstra algorithm / A* ( A-star ) algorithm Static or dynamic Static: Norminal travel time (speed limit) Dynamic: Current travel times, updated in real-time (Historic: Historical median travel times) (Predicted: From our travel time prediction method; for longer routes only) Web-based user interface and the map server

8 Web-based user interface Similar to commercial webbased routing services such as Mapquest It receives the street addresses of O-D from the user and displays the best route found by the algorithm on the map and the turn-by-turn directions. Also shows nominal, historical average and predicted travel times, along with other information such as daily travel time predictions over the route and incidents

9 Benefit of Dynamic Route Guidance How much time a traveler saves using the dynamic route guidance system compared to when he/she uses the static one? Study the LA freeway network Computed shorted paths for each of them for each departure time every 5-minute from 5 AM to 11 PM on October 4, 2004, using both static and dynamic route guidance systems

10 Los Angeles (Caltrans District 7) freeway network Screenshot from PeMS website Each dot corresponds to a loop detector station and the color of the dots show the current speed at each location

11 The simplified Los Angeles freeway network The directed graph with 21 nodes and 64 directed edges. There are 420 possible O-Ds

12 The last plot shows an average of 30~60% reduction in travel time for trips that last an hour or longer

13 Summary A prototype web-based dynamic route guidance system we developed Uses the real time traffic information and historical data archive from PeMS A sophisticated travel time prediction algorithm to report a realistic estimate of the trip duration, reflecting the possible change in the traffic condition during the trip. The benefit of the dynamic route guidance system compared to the static route guidance system. An average of 30~60% reduction in travel time for trips that last an hour or longer. Even when there s no alternative route, the realistic travel times reported by the system benefit drivers in trip planning. To summarize, we developed a robust and mature web-based dynamic route guidance system that makes intelligent use of real-time and historical data in PeMS. Combined with realistic travel times computed using a travel time prediction algorithm, the system can benefit travelers in various ways.

14 References PeMS Website, Accessed February 1, Mapquest Website, Accessed February 1, Pattanamekar, P., Park, D., Rilett, L.R., Lee, J. and Lee, C. (2003) Dynamic and stochastic shortest path in transportation networks with two components of travel time uncertainty," Transportation Research Part C, 11, Dijkstra, E.W. (1959) A note on Two Problems in Connexion with Graphs, Numerische mathematik, 1, Sedgewick, R. and Vitter, J. S. (1986) Shortest paths in Euclidean space, Algorithmica, 1(1):31. Kwon, J. and Petty, K. (2005). A Travel Time Prediction Algorithm Scalable to Freeway Networks with Many Nodes with Arbitrary Travel Routes, Presented at 84th TRB Annual Meeting at Washington D.C., Forthcoming in Transportation Research Record.