Objective Reduction on Many-Objective Traffic Lights Signaling Optimization

Traffic Lights Signaling Optimization consists in optimizing traffic-light cycle using optimization algorithms. In this context, a traffic simulator can represent different road intersections and traffic lights and can calculate traffic quality measures. A solution to this problem can be encoded in a vector representing the time of each traffic light on the simulator and the objective functions are the traffic quality metrics. Since there are several metrics this problem can be defined as a many-objective optimization problem. In spite of the existence of several objective functions, the majority of the related work selects a small set, often two. This paper proposes a many-objective optimization framework based on objective reduction to solve Traffic lights Signaling Optimization Problem. Here, twelve objective functions are used. To deal with the high number of metrics, objective reduction techniques are applied along a multi-objective evolutionary algorithm. An experimental set is conducted to analyze if it possible to reduce the number of objective functions on the Traffic lights Signaling Optimization Problem and if this reduction enhances the performance of the optimization algorithm.