Published: 2009 Mai
Buchtitel: Proceedings of mobil.TUM 2009 - International Scientific Conference on Mobility and Transport
Verlag: TU München
Evolutionary algorithms are optimisation heuristics that are inspired by biological evolution. They are relatively easy to comprehend and can be applied to any problem where a fitness function for rating candidate solutions is available. Therefore, evolutionary algorithms have been successfully applied to a wide range of real-world problems since their development in the 1960s. Since several years, their application domain also includes the optimisation of traffic signal systems. Here, the challenges are the often time-consuming and noisy fitness evaluations that are in many cases based on stochastic traffic simulations. The resulting time requirements make the use of evolutionary algorithms a challenging task especially in on-line scenarios where the traffic signal system has to be continuously adapted to changing traffic demands. This paper presents a structured overview of evolutionary algorithm applications in traffic signal optimisation. Different (off- and on-line) scenarios are presented and techniques for reducing their time requirements are discussed. Furthermore, multi-objective evolutionary algorithms that simultaneously treat several (contradicting) objectives are introduced.