A neuro-fuzzy system for the prediction of the vehicle traffic flow

In this paper, we propose a fuzzy system to control vehicle traffic flows on a street network. At a given point of the street network, data are collected by a peripheral unit equipped with infrared sensors. Row data are sent by the GSM/GPRS network to a centralized data processing server, where a simple set of fuzzy rules is employed to classify the row data samples into three flow states corresponding to flowing, intense and congested conditions. The core of the system is constituted by a neuro-fuzzy system, which is used to predict the time series constituted by the fuzzy membership of traffic measures to the three predefined flow states. We report the results concerning the comparison tests we carried out using an ANFIS network synthesized by a hyperplane clustering procedure and some well-known prediction techniques used as benchmarks.