Optimisation of fixed-time signal timings for area traffic control using genetic algorithm

Abstract

In this paper, a genetic algorithm is developed to optimise the fixed-time signal timings for area traffic control. The traffic model from TRANSYT is employed to evaluate the performance of a signal-controlled network. The encoding method, reproduction, crossover, mutation and fitness scale in the genetic algorithm are discussed and the detailed implementation of them in the optimisation process is also given. The advantages of the proposed genetic algorithm over the conventional hill-climbing method in TRANSYT are that the probabilistic nature of the genetic algorithm enables a near global optimum to be determined, since the search process covers more extensively the feasible region of the problem; and in contrast to the signal oplimiser in TRANSYT the proposed methodology enables the common cycle time, offsets, green splits and stage ordering to be optimised together in a unified manner.