Traffic Signal Optimization with Estimation of Distribution Algorithms and VISSIM-MATLAB Integrated Simulation Platform

Traffic signal optimization plays a crucial role in improving the service ability of traffic networks in urban areas. With the traffic network getting more and more complex, there have been increasing research interests in employing intelligent algorithms to find proper settings for traffic signals. As a special type of evolutionary algorithms, estimation of distribution algorithms (EDAs) possess strong optimization ability but have seldom been used in traffic signal optimization. In this paper, two efficient variants of continuous EDAs, namely EDA with variance enlargement strategy (EDA(ve)) and EDA with variable-width histogram model (EDA-VWH), are modified and adopted as optimizers to find proper traffic signal cycles in an actual urban area with multiple intersections. The performances of the two resultant algorithms, i.e. modified EDAve (mEDA(ve)) and modified EDA-VWH (mEDA-VWH), are comprehensively studied through a VISSIM-MATLAB integrated simulation platform, which could provide a convenient and close-to-reality simulation environment. The simulation results showed that mEDAve and mEDA-VWH could effectively reduce the mean delay time of all vehicles under different traffic conditions. In comparison with four other algorithms, including genetic algorithm, particle swarm optimization, differential evolution and random search method, the two modified EDAs also achieved competitive results.