A NEURAL NETWORK BASED APPROACH FOR PREDICTING NETWORK TRAFFIC CONDITIONS

This paper presents a new approach for the short-term predictions of network traffic conditions, which include near future traffic flow, travel time, and delay time in various locations of a traffic network. The proposed approach employs neural network and microscopic simulation techniques to perform the prediction on real-time basis. Many ITS user services need short-term traffic condition predictions on real-time basis to support full scale deployment and to improve the quality of the user services. In current ITS practices, short-term predictions have not yet been proposed or implemented. The lack of the predictions may cause technical difficulties in implementing and improving ITS user services. A new approach is proposed in this paper, which uses a quite different methodology in solving the problems in required computing power and real-time link data. The approach employs artificial neural network (ANN) and microscopic simulation technique to build a framework, which can take real-time traffic data on partial links of the study areas and output the traffic condition predictions on real-time basis using available computing resources such as PC or work station.

Language

  • English

Media Info

  • Features: Figures; References;
  • Pagination: p. 507-516

Subject/Index Terms

Filing Info

  • Accession Number: 00724769
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Aug 16 1996 12:00AM