Optimization of Zero-Order TSK-Type Fuzzy System Using Enhanced Particle Swarm Optimizer with Dynamic Mutation and Special Initialization

Abstract

This paper proposes an enhanced particle swarm optimization with dynamic mutation and special initialization, the EPSODM-I, for improving the accuracy of zero-order TSK-type fuzzy systems design. The EPSODM-I can be regarded as a new population-based evolutionary optimization algorithm. Unlike the generic initialization used in most popular population-based algorithms, the EPSODM-I applies a proposed special initialization method to generate the initial PSO particles for fuzzy system design. The generated initial PSO particles are iteratively improved by a new approach incorporating the dynamic mutation into the existing PSO to provide more diverse search directions. Application examples of the zero-order fuzzy system designs for the tracking control of nonlinear dynamic systems have been simulated to validate the proposed algorithm. In terms of tracking errors, performance comparisons with the typical PSO and different advanced PSO variants verify the superiority of the proposed algorithm. The effects on the convergence rate and optimization accuracy yielded by the proposed special initialization and dynamic mutation have also been discussed and verified crossly in the simulation results.