CPG parameter search for a biomimetic robotic fish based on particle swarm optimization

Abstract

This paper addresses the parameter search issue of a Central Pattern Generator (CPG) governed fishlike swimming. Since the CPG parameters involving amplitudes, frequencies, and phase lags are closely related to the propulsive performance, an idea optimizing the CPG characteristic parameters for the maximum propulsive speed is formed and implemented. Specifically, a dynamic model of robotic fish swimming using Kane's method is developed to guide the primary parameter search. A particle swarm optimization (PSO) algorithm is further employed to optimize the CPG parameters for an enhanced performance. Numerical simulations and robotic experiments superior to previously published results are finally given, validating the effectiveness of the PSO-based search scheme.