Author(s)

G. Ghiani, D. Gullì, F. Mari, R. Simino & R. Trunfio

Abstract

The estimation and prediction of link travel times in a road traffic network are critical for many intelligent transportation system (ITS) applications, such as the route guidance system (RGS), advanced traveller information system (ATIS) and freeway traffic management system (FTMS). These systems are adopted to help individual drivers to identify optimal routes based on real-time information on current traffic conditions. The identification of the optimal routes is particularly important for trips where the travel time is relatively long and where it is unlikely that the current travel time will remain stable. In addition, weather conditions and road type are both parameters that influence the travel time of a specific link, and they need to be included in the forecasting process. The aim of our study is to propose a new system for travel time forecasting based on a multilayer feedforward neural network. Both historical and real-time data (which can be provided by loop detectors and sensors positioned along the roads) are inputs for the neural network that returns the short-term travel time needed to traverse the road section that it is related to. Data used to train and test the neural network have been generated using a simulator that is influenced by deterministic (e.g. road type) and stochastic (e.g. weather, visibility) parameters. Keywords: travel time forecasting, neural network. 1 Introduction The problem of travel time prediction has received over the years more and more attention from researchers and many methods have been proposed. The main idea of travel time forecasting is based on the fact that traffic behaviours possess both partially deterministic and partially chaotic properties. Good forecasting

Keywords

travel time forecasting, neural network.