Freeway Traffic Control based on Neural Network Estimation

Due to their characteristics, freeways appear to be ideal sites for testing new traffic regulation strategies, based on continuously improved information systems. In this framework, a particular consideration has recently been devoted to the application of neural networks (NNs) to freeway supervision and control. In this paper, the solution to a "classical" traffic control problem is tackled with. The traffic state variables given as inputs to such a problem can be computed via a macroscopic traffic model, thus requiring costly and complicated varying parameter identification, or via a significantly simpler NN filtering approach. While the second approach does not require to identify the parameters every time they change, it will be shown that the performances of such two computation procedures are comparable. Copyright IFAC 2005