Real-Time Traffic Signal Control with Dynamic Evolutionary Computation

Nowadays real-time traffic signal control is a crucial issue with potential benefits in the fields of traffic control, environmental pollution, and energy utilization. In the literature, few related studies have been done with dynamic evolutionary algorithms. In this paper, we proposed a strategy using Collaborative Evolutionary-Swarm Optimization (CESO), which is able to track time-varying optimal solutions effectively. We use the simulator of urban mobility (SUMO), a popular traffic simulator to generate traffic flows. A grid traffic network is designed with several scenarios to simulate changes of traffic flows captured by traffic monitors. We test different traffic changes in the network using the proposed strategy and compare its performance with a traditional evolutionary algorithm. Experimental results show that our algorithm can obtain promising configuration of traffic light cycles and reduce the average delay time of all vehicles in various scenarios.