Global localization of Monte Carlo localization based on multi-objective particle swarm optimization

Premature convergence often happens when a Monte Carlo localization (MCL) algorithm tries to localize a robot under highly symmetrical environments. In this paper, we propose a novel method of solving such problem for global localization by incorporating a multi-objective evolutionary approach to resample particles with two objectives, including particle weights and population distribution. By employing a multi-objective particle swarm optimization (MOPSO), our approach is capable of enhancing the exploration ability to improve population diversity while maintaining convergence quality to successfully localize the global optima. Simulation results have confirmed that localization performance using the proposed approach is significantly improved.