Probabilistic Urban Link Travel Time Estimation Model Using Large-Scale Taxi Trip Data

Accurate estimation and prediction of urban link travel times are critical for urban traffic operations and management. This paper develops a probabilistic mixture model to estimate urban link travel times using large-scale taxi trip data. Unlike typical Global Positioning System (GPS) trace data, the taxi trip data used in this study provide limited trip level information, which contain only the locations of origin and destination, travel times and distances, etc. The focus of this study is to develop a robust probabilistic link travel time estimation model and demonstrate the feasibility of estimating network conditions from large-scale origin-destination (OD) level trip data. In the model, the path taken by a taxi is considered as latent and taxi drivers’ route choice behaviors are modeled using a multinomial logit distribution. The observed path data given the possible paths set and the link travel time parameters can be thus characterized using a mixture distribution. An efficient solution approach based on expectation-optimization (EM) algorithm is proposed to solve the problem. The model is tested on estimating the distribution of the link travel times for 30 min time periods using real-world data from Midtown Manhattan, New York City. Robust estimation results are obtained owing to the adoption of a probabilistic framework.

• Supplemental Notes:
  • This paper was sponsored by TRB committee ABJ30 Urban Transportation Data and Information Systems. Alternate title: A Probabilistic Urban Link Travel Time Estimation Model Using Large-Scale Taxi Trip Data.
• Corporate Authors:

  Transportation Research Board

  500 Fifth Street, NW
  Washington, DC  United States  20001
• Authors:
  • Zhan, Xianyuan
  • Ukkusuri, Satish V
• Conference:
  • Transportation Research Board 94th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2015-1-11 to 2015-1-15
• Date: 2015

Language

• English

Media Info

• Media Type: Digital/other
• Features: Appendices; Figures; References; Tables;
• Pagination: 18p
• Monograph Title: TRB 94th Annual Meeting Compendium of Papers

Subject/Index Terms

• TRT Terms: Algorithms; Estimating; Multinomial logits; Origin and destination; Route choice; Taxicabs; Travel time; Urban travel
• Uncontrolled Terms: Probabilistic models; Trip data
• Geographic Terms: Manhattan (New York, New York)
• Subject Areas: Data and Information Technology; Planning and Forecasting; Public Transportation; I72: Traffic and Transport Planning;

Filing Info

• Accession Number: 01553700
• Record Type: Publication
• Report/Paper Numbers: 15-4054
• Files: TRIS, TRB, ATRI
• Created Date: Feb 19 2015 12:46PM