This paper describes the development of a relatively new-generation of algorithms for freeway incident detection using artificial neural networks (ANNs). These new ANN models have the potential to provide faster and more reliable incident detection times and fault-tolerant operation while being easy to implement on existing and new hardware platforms. For model development, two sources of data were collected from an 8.5 km segment of the Tullamarine Freeway in Melbourne, Australia. These comprised speed, flow and occupancy data from dual-loop detector stations and an incident log showing the approximate time of incidents. This resulted in a database comprising a set of incidents of varying severity and under a variety of traffic conditions at different locations of the freeway. These data were used to train the ANN models. These models classify the traffic conditions into either incident or incident-free states. An incident is detected when the traffic state changes from incident-free to incident conditions. The results obtained from a number of model runs demonstrate that 'real world' data can be used to train ANN incident detection models. (a) For the covering entry of this conference, see IRRD abstract 868345.


  • English

Media Info

  • Features: References;
  • Pagination: p. 123-40

Subject/Index Terms

Filing Info

  • Accession Number: 00722229
  • Record Type: Publication
  • Source Agency: ARRB
  • ISBN: 0-86910-663-5
  • Files: ITRD, ATRI
  • Created Date: Jun 28 1996 12:00AM