Traffic Lights Management Using Belief Propagation

Abstract

There is a considerable interest in the investigation of traffic control in road networks. Road congestion is now part of daily life in urban environment. As cities develop, traffic congestion only worsens, and constructing more roads does not always resolve the problem. One of the approaches is to improve traffic lights management and make them adaptive to the traffic conditions. Using the traffic information, one aims to estimate the best traffic lights configuration for each junction. Using probabilistic methods, message passing techniques [1] can be applied to a road network, represented by a sparse graph, such that each junction shares traffic and traffic light control information with neighbouring junctions. It is the aim of the project to devise a belief propagation algorithm for efficient traffic lights management of a given road network according to the current traffic information. Two inherent cost/success measures will be employed based on equating traffic flow in all roads and on limiting traffic density per road below a fraction of its capacity. Simulations were carried out for both cases, and the algorithm gives good results. Depending on the cost function used, the algorithm manages to balance the loads or reduce the number of congested roads.