Distributed Particle Swarm Optimization for Limited Time Adaptation in Autonomous Robots

摘要: Evaluative techniques offer a tremendous potential for on-line controller design. However, when the optimization space is large and performance metric noisy, time needed to properly evaluate candidate solutions becomes prohibitively and, as consequence, overall adaptation process extremely consuming. Distributing reduces required increases robustness failure of individual agents. In this paper, we analyze role four algorithmic parameters that determine total evaluation in distributed implementation Particle Swarm Optimization algorithm. For multi-robot obstacle avoidance case study, explore simulation lower boundaries these with goal reducing so it feasible implement within limited amount determined by robots’ energy autonomy. We show each parameter has different impact on final fitness propose some guidelines choosing real robot implementations.