Urban Traffic Management; the Viability of Short Term Congestion Forecasting Using Artificial Neural Networks



Urban Traffic Management; the Viability of Short Term Congestion Forecasting Using Artificial Neural Networks

Authors

LYONS G, McDONALD M and HOUNSELL N, University of Southampton, WILLIAMS B and CHEESE J, Smith Systems Engineering Ltd, and RADIA B, Department of Transport, UK

Description

Artificial neural networks (ANNs) are now widely recognised as a facet of artificial intelligence. ANNs represent a set of modelling techniques which are distinctly different from more conventional approaches. This consideration, in conjunction with some

Abstract

Artificial neural networks (ANNs) are now widely recognised as a facet of artificial intelligence. ANNs represent a set of modelling techniques which are distinctly different from more conventional approaches. This consideration, in conjunction with some notable ANN modelling successes, has prompted researchers in a broad and diverse range of fields (including transportation) to assess the potential of ANNs as an alternative to existing techniques or, indeed, as a solution to problems formerly lacking an appropriate modelling technique.

Rising traffic levels and consequently an increase in the frequency and severity of congestion have prompted considerable investment in the development of traffic management techniques. Traffic management, particularly of an urban road network, relies increasingly on knowledge of the road network traffic status obtained from a growing infrastructure of network monitoring equipment. Consequently a progressively data rich environment has evolved to support traffic management system(s)'. Traffic flows in an urban network reflect a series of underlying highly non- linear relationships. ANN techniques are notable for their use in addressing non-linear problems in data rich environments.

Set against this background, the Department of Transport commissioned the Transportation Research Group at the University of Southampton, together with Smith System Engineering Limited, to assess the potential application of ANN techniques to urban traffic control (UTC) and in particular to assess the viability of using an ANN technique to forecast the onset of urban congestion.

This paper describes this initial study which was conducted over a six month period. A comprehensive review of key issues considered ANN techniques, their applications in transportation and traffic control, potential data sources and congestion definition and management. Based on the findings of the review, a demonstrator application was defined with the intention of exemplifying the potential of a UTC application of ANNs. The demonstrator application, in fact, performed a much broader role, providing a focus to the study and thereby acting as a catalyst for generation of further understanding of the relevant issues and implications. The paper provides an elementary level of technical detail and places emphasis on consideration of the issues and implications that emerged from the research.

Publisher

Association for European Transport