Fuzzy Traffic Signal Control and a New Inference Method ! Maximal Fuzzy Similarity

New methods, like fuzzy logic, are coming into the field of adaptive traffic signal control. Development of the fuzzy control can roughly be divided into two research approaches: development of fuzzy control functions, and development of fuzzy inference methods. Both approaches are discussed in this paper. First, a lately developed fuzzy inference method, called maximal fuzzy similarity, is introduced. Second, two fuzzy traffic signal control functions, phase selector and green extender, are presented and their performance is evaluated by simulations. Third, the applicability of the maximal fuzzy similarity inference method in traffic control systems is compared to the traditional Mamdani inference method. The comparison is made using them separately in the above mentioned control functions. In the simulations, the phase selector function improved significantly the control performance, while the fuzzy green extender worked better than the non-fuzzy control only with high volumes. The fuzzy sets and inference seem to have meaning only when the input values (volumes) are high enough. The main difference between the tested inference methods was the fact, that the defuzzification method of Mamdani inference, center of area, leads to more compromising control decisions than the similarity method. Instead of combining the outputs of all the control rules, the similarity inference launches the action of the rule with highest similarity to the input values and ignores the output values of other rules. Thus, the Mamdani method is slightly more appropriate to conditions, where compromises are needed, while the similarity method with more extreme control actions has better performance with conditions requiring more radical adjustments to the control.