Real-Time Hazardous Traffic Condition Warning System: Framework and Evaluation

This article discusses a proposed freeway warning information system that could be coupled with an ability to identify hazardous traffic conditions based on probabilistic Bayesian modeling. A probabilistic neural network (PNN) is used in this Bayesian model, using probability density functions (PDFs), to operate an entirely parallel non-feedback network that does not require retraining used in error back propagation like multilayer feedforward (MLF) networks do. This PNN method was implemented in the Hayward, California I-880 freeway project to collect real-time data from a strip of freeway about 9.2 miles long. Flow, occupancy, and speed information was collected by sensors at ten second intervals- then, upon the statistical compilation of this data, it was fed into the PNN algorithm in order to train it to determine threshold levels of vehicle safety variables. Researchers found that, while the threshold indication could be indicated through one of two PNN models they had established, a combinatory approach toward the salient features of each network would be the optimal model for them to work with. Researchers also predict that such warning modeling systems will be crucial for future intelligent transportation system technologies (ITS).

  • Availability:
  • Authors:
    • Oh, Cheol
    • Oh, Jun-Seok
    • Ritchie, Stephen G
  • Publication Date: 2005-9


  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01010957
  • Record Type: Publication
  • Source Agency: UC Berkeley Transportation Library
  • Files: BTRIS, TRIS
  • Created Date: Nov 30 2005 12:14PM