A PSO-Optimized Fixed and a PSO-Optimized Neural Network-Adaptive Traffic Signal Controllers for Traffic Improvement in Santo Domingo, Dominican Republic

Abstract

Satisfying the mobility demand is one of the biggest concerns arising with the increase of urban population. With many people in the road network, traffic congestions are present in most of the cities in the world. The Distrito Nacional in Santo Domingo, capital city of Dominican Republic, is a notorious example of this phenomenon. Unfortunately, all the efforts to improve traffic experience there have had little success. With this work, two models have been developed using Particle Swarm Optimization (PSO): a PSO-Optimized Fixed Traffic Signal Control (PSO-FTSC) and a PSO-Optimized Neural Network-Adaptive Traffic Signal Control (PSO-NN-ATSC) that uses 4 neural networks to predict phase times. The intersection of 27 de Febrero Avenue corner with Winston Churchill Avenue was simulated using Simulation of Urban Mobility (SUMO), minimizing the time loss per vehicle during optimization. These models, PSO-FTSC and PSO-NN-ATSC, present reductions of 17% and 24% of mean time loss, respectively. These promising results may lead to a decrease of fuel consumption, reducing the consequent air pollution, as well as to an improvement of businesses and people’s productivity and mental health.