Optimization of traffic lights operation using network load data

Abstract

Control of traffic lights in a road network of a large city is an extremely complicated task. However, the efficient use of network capacities is impossible without adjusting traffic lights operation cycles. To date, a lot of methods have been developed to adjust the operation of traffic lights based on classical transportation models. However, classical models of flow distribution are not sensitive to local effects taking place at intersections of a road network. This article shows that artificial neural networks and randomized algorithms of stochastic approximation allow building systems for traffic lights operation control that take into account various non-linear stochastic relations between locally observed network loads. The article describes a method to get such control and presents the results of testing the approach through the example of a test transportation network.