| Summary: | 碩士 === 國立嘉義大學 === 電機工程學系研究所 === 104 === This thesis proposed a new metaheuristic population-based evolutionary optimization algorithms, mutation-aided elite continuous ant colony optimization (MECACO), for the design of TSK-type fuzzy neural network (FNN) and TSK-type recurrent fuzzy network (TRFN). The basic principle of MECACO is a stochastic search algorithm, which combines a new designed elites-based continuous ACO with the mutation technique employing the dynamic mutation probability to exploit and explore the solutions globally at the same time. The MECACO was used to the reinforcement learning of the FNN and TRFN for the tracking control of the nonlinear dynamic plants, chaos prediction, and system identification to show its effectiveness. The performance superiority of the MECACO to different continuous ACO-based algorithms is demonstrated through simulation examples of the fuzzy systems design.
|
|---|
