Online Learning Solutions for Freeway Travel Time Prediction

In order to improve travel time information to travelers, as this information can improve individual driving behaviors and choice behaviors in addition to collective congestion on roadways, the author of this article presents a means of implementing a data-driven approach using state-space neural networks (SSNN). SSNNs can be implemented in a variety of ways, with information being distributed to individual drivers via variable message signs (VMS). It is emphasized that an SSNN to be implemented for travel time information must take account of variability in road conditions and thus be highly adaptive. Additionally, travel-times are contingent upon the realization of the trips in progress, which the travel time information itself can alter. Using an extended Kalman filtering (EKF) technique, the author implements an online learning approach that improves upon delayed approaches that only take account of realized travel times.

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01090989
  • Record Type: Publication
  • Source Agency: UC Berkeley Transportation Library
  • Files: BTRIS, TRIS
  • Created Date: Mar 13 2008 1:54PM