This paper describes an investigation of the feasibility of optimising a prototype fuzzy logic signal controller with respect to several criteria simultaneously. The simple building blocks of a fuzzy control system are fuzzy sets, which capture significant categories of input and output values, and rulebases, which describe relationships between inputs and outputs. The prototype controller uses a competitive technique, in which a weight for each phase (or signal group) at the junction is derived each second by several fuzzy control modules. The weights are derived from traffic count data from sensors and signal timing data and data about neighbouring junctions. They are combined to give a weight for each possible stage, and used as a basis for a decision to change stage or retain the current stage. It was decided to use the recently developed Multiobjective Genetic Algorithm (MOGA) approach, which allows a range of optimal candidate solutions to be found, rather than impose an arbitrary weighting of the various criteria to be optimised to lead to a single solution. The paper describes the evaluation of a traffic signal controller using this method, and considers the traffic simulations used, the optimisation of the controller, and the results of the study. The controller's sensitivity has been demonstrated.
Abstract