Localização de robôs móveis utilizando o novo algoritmo de Morcegos baseado em Líderes

Abstract

A complex stochastic problem of essential importance in mobile robot applications is the location of the robot. This can be treated as an optimization problem, due to its characteristic of presenting different possible solutions (local minimums). The method proposed in this work is based on the Bats Algorithm, which presented better results for the problem of global localization and robot sequestration when compared to other methods in the literature. We presented favorable results for both global localization, with a speed as skillful as the compared algorithms, as significant results in a robot sequestration situation. The proposed name for the modification of the classic bats algorithm was Leader-Based Bat Algorithm (LBBA), because it has the characteristic of simulating actions of group leaders, which influence others less skilled. The consequence of using more than one leader implies a search, the best position, the most diversified on the map, dealing better with scenarios that present many ambiguities during the localization. Tests were performed in different scenarios (besides the test on a real map). The bio - inspired algorithm of bats in its classical form (BA) was compared with other algorithms (particle swarm optimization - PSO and Particle Filter - FP) and obtained good results. These inspired the proposal of a variant in the bats algorithm capable of dealing with more complex situations of location where the BA presented difficulties, the LBBA obtained remarkable results when compared to Monte Carlo Adaptive Location (AMCL).