The Beta distributed PSO, β-PSO, with application to Inverse Kinematics

摘要

this paper introduced a new Particle swarm optimization, PSO, variant where Beta profiles are used to manage exploration and exploitation behaviors of the swarm. A key issue in PSO is that it is missing a clear separation between the exploration behavior and the exploitation behavior of the swarm. In Beta-PSO exploration/ exploitations phases are clearly identified and particles evolve differently in each phase. A couple of beta distributions are used to control both phases. In the exploration phase, particles evolve based on a pseudo-normal beta profile, while in the exploitation phase a decreasing exponential beta distribution is used. The proposed method is applied to solve the inverse kinematic problem of a 6 axes generic robot arm. Results showed that the quadratic error of Beta-PSO was about 5.5 e-17, while QPSO solutions were at the level of 5.6e-l0; SSA returned 6.6 e-8 error and classical PSO was at about 1.7 e-3. Those results were confirmed with the Wilcoxon non parametric comparison which clearly showed that Beta-PSO is better than the classical PSO, quantum PSO, and SSA for this application.