Multi-Objective Vibration-Based Particle-Swarm-Optimized Fuzzy Controller With Application to Boundary-Following of Mobile-Robot Simulation Environment

摘要

This paper presents a multi-objective vibration-based particle-swarm-optimization (MO-VBPSO) algorithm with enhanced exploration ability and convergence performance, for training fuzzy-controller (FC) to achieve robot control. The MO-VBPSO applies a reference point-based leader selection schema that assigns leaders for MO-PSOs’ searching optimal parameters of the FC. Besides, the MO-VBPSO framework is integrated with a vibration factor to strengthen the exploration ability for resolving the local minima issue, which is inspired by the amplitude of the Firework Algorithm (FWA). The evaluation of MO-VBPSO focuses on the effect of the vibration factor by applying it to training a mobile robot in a simulation environment. The evaluation results are discussed concerning exploration ability, convergence performance, and performance stability. Experimental results reveal that the proposed MO- VBPSO lifts the performance of robot training significantly.