The role of traffic signals is evolving to include people-oriented as well as vehicle-oriented objectives. Future signal control systems will need to handle a wide range of input types as improved models (for example, cyclist and pedestrian behaviour) and other objective functions (for example, environmental concerns) become available. Fuzzy logic systems can be built to handle existing vehicle-based objectives and are readily extendable using the hierarchical approach. Some fuzzy logic traffic signal control systems based on vehicle-oriented objectives have been built and tested against simulated models since 1977. Two of the more recent efforts produced at the Transport Operations Research Group, University of Newcastle upon Tyne, were designed to incorporate policy-sensitive objectives in the future but tuning the systems proved time-consuming and tedious. Automatic optimisation techniques were sought and genetic algorithms were proposed as a possibility. This paper discusses the elements of fuzzy logic systems and pinpoints those that could be optimisable using genetic algorithms. Methods applicable to optimising fuzzy logic traffic signal control systems offline are suggested. (A)
Abstract