Dealing with a problematic roundabout by optimizing a traffic light system through evolutionary computation

Abstract

The Padre Anchieta roundabout (Tenerife) takes on a substantial amount of traffic which, at peak times, often causes significant traffic jams. With the aim of alleviating traffic congestion, this paper analyses the possibility of installing a set of traffic lights in the roundabout. The duration of the phases of traffic lights are optimized through a genetic algorithm. SUMO, an open source traffic simulator, is used to evaluate the traffic of the area by considering real traffic data. Seven different scenarios are evaluated with respect to the roundabout, three of them without traffic lights and the other four with them. Simultaneously, the number of pedestrians, and the particular location of the traffic lights, are modified in each of the aforementioned scenarios, when applicable. To determine which parameters of the evolutionary algorithm provided the best results, a previous parameter setting study based on a statistical comparison procedure was performed. Particularly, the crossover operator and the population size, were analyzed. From the results obtained in the experimental assessment, we conclude that the use of an optimized traffic light system would not improve traffic flow in the roundabout. Another important conclusion is that the larger the number of pedestrians, the slower the traffic flow.