Development of discrete differential evolution algorithm for solving traffic light scheduling problem

Abstract

This project focus on researching the performance of the discrete differential evolution algorithm applied in the pedestrian-vehicle mixed-flow network scheduling problem. Given the pedestrian and vehicle flow mixed traffic model that describes both pedestrian behavior and vehicle behavior on the macroscale, the scheduling problem turns into a mathematic model which could be optimized by various algorithms. Based on the model structure, the discrete differential evolution algorithm is developed to find the optimal solution in discrete space. Inspired by the traditional continuous differential evolution algorithm, two discrete mutation operators is proposed to fit this model. To help improve the performance of the mutation operator, a greedy-local-search operator is proposed by summarizing the feature of several local operators. The combined discrete algorithm merged the advantages of various operators, thereby, it has a more ideal performance in the traffic light scheduling problem.