Adaptive Fuzzy Traffic Signal Control Based On Genetic Algorithm

Adaptive fuzzy control is a research field, which is composed of the theory of fuzzycontrol and adaptive control. Adaptive fuzzy control has recently attracted increasing attentionfrom control circles, because the system of adaptive fuzzy control can not only improvesystem performance by modifying own rules but also has quick convergence speed androbust.As the fuzzy rules and the membership functions of the traditional Fuzzy LogicController are always given by human experiences beforehand and once decided neverchanging during the whole control period, the coming problem is that when the trafficcondition changes too much, especially, the vehicles flow of different directions arediscrepant greatly, the control effect becomes worse. So the Genetic Algorithm is used toreal-timely optimize the FLC's fuzzy rules and membership functions to form self-adaptiveFLC because the Genetic Algorithm has got the strong searching ability, without knowingthe mechanism of the object and only driven by the fitness function.The background of the traffic signal control in ITS and the research work of optimal FLCused in the intersection signal control is summed up in this thesis. The designation method ofall kinds of FLCs' is Classified, and the designing details is explained. The principle ofGenetic algorithms, the essence of three operators and the reason of local best are analyzed inthis thesis. Question of GA optimizing respectively the fuzzy rules and membership functionsof the traditional FLC using in isolate intersection signal control is discussed in this thesis.Aiming at the difference between the urban traffic network and the isolated intersection, thedouble FLC's system, and real-timed optimizing the sequential membership functions of theinputs and output of the Secondary FLC is designed in this thesis. the coding scheme,fitness function, GA operator, etc details are explained. The effectiveness is proved by thesimulation result of MATLAB.At the end of this thesis, as the direction of future research, a new immune geneticalgorithm which has the processes of immune memory and antibodies concentrationadjustment is proposed, the antibody variety is guaranteed, the algorithm gets in local minimaand precocity convergence to be avoided.