Modified Symbiotic Evolutionary Learning for Type-2 Fuzzy System

This paper proposes a new modified symbiotic evolutionary learning method (SET2) for fine-tuning the parameters of a type-2 fuzzy system. The paper uses a group-based symbiotic evolutionary approach. Instead of evolving the individual rules and assigning the fitness values for each rule, the membership function parameters and the rules are evolved as individual groups. This method improves the exploration of the solution space, and optimizes the rule, base count and the most influencing input parameters in each rule which is difficult to obtain in conventional symbiotic evolutionary learning approach. The paper also proposes a method to obtain the similarity measure of the membership functions. In order to demonstrate the efficiency of the proposed learning method, it is deployed in a simulated urban traffic network to obtain the optimal signal timings in a distributed manner. The proposed SET2 signal control was tested for various traffic conditions experienced in a section of the Central Business District of Singapore. A comparison of the average time delay and speed of vehicles indicated that SET2 signal control performed significantly better than the benchmark signal controls such as GLIDE, hierarchical multiagent system, and geometric fuzzy multiagent system. The test bed also showcased the performance of the modified learning method in a distributed environment.