Planejamento inteligente de trajetórias para navegação autônoma e cooperativa de múltiplos robôs

Abstract

Robotic applications are highlighted in several areas, especially those that consider multi-robot systems, in which multiple mobile robots are able to develop a task in an autonomous and cooperative way. In this context, intelligent systems are valid tools to provide a navigation and coordination of a multi-robot system. This work has as main objective to present solutions to problems that require the path planning for multiple robots, defined from an environment that contains points of obligatory passage (targets) and static and dynamic obstacles. In this environment, the multi-robot system must visit all targets in a cooperative way and finish the navigation in a pre-defined location. Initially, the knowledge of the environment is limited only on the targets' position (obstacles are unknown), and then a hybrid path planner is proposed to compute the paths to be navigated by the robots from that previous knowledge (off-line step) and handle unexpected problems during navigation (on-line step, individually performed by each robot), such as obstacle avoidance and loss of localization. The hybrid planner is modeled through a multiple traveling salesperson problem adaptation, and solved with genetic algorithms and other cooperative heuristics. In a second step, the previous knowledge about the environment consists only on the initial and final positions of the robots. In this case, robots should explore the environment to locate and visit the targets, while avoiding obstacles, until they reach their final positions. To do this, an on-line and bio-inspired strategy on stigmergic behavior from artificial pheromones is proposed to provide a cooperative and distributed navigation. The robots considered in this work are homogeneous, independent, have limited communication skills and are equipped with fuzzy controllers for their locomotion. Results from simulation are presented for the planners validation and a implementation with real robots in an experimental environment confirms the feasibility of the proposed strategies.