Travel times on urban controlled links : a neural network approach.

For a long time restricted to motorways, road traffic information has emerged nowadays as an increasing need on urban networks. This paper presents concrete results from research aiming at testing the feasibility of an on-line travel time display on controlled urban arteries in Paris. The methodological approach is based on neural networks. Data necessary for the calibration of the networks are extracted from real surveys. They gather volume and occupancy as input from measurement stations on the urban itineraries, and travel time values as reference output, measured by floating-car surveys. The results obtained are, in some cases, quite promising. They underline the importance of the number of stations and traffic signals on the trips. The feasibility of an on-line travel time display in urban areas seems therefore established.