Controle descentralizado com desvio de obstáculos para uma formação líder-seguidor de robôs móveis

Abstract

This thesis describes a decentralized control scheme proposed for guiding the navigation of a leader-follower formation of mobile robots, including the possibility that the whole formation avoid obstacles. A laser scanner is here adopted to provide the information necessary to accomplish tasks like positioning and displacement of the follower robot with respect to the leader one. Modifications in the tangential escape algorithm (implemented on the leader robot) are presented and experimentally tested for avoiding concave obstacles configurations. For semi-structured environments, a leader searching strategy is developed using the range measurements provided by the laser scanner mounted on the follower robot. In this leader-follower formation, there is no information sharing, emphasizing the decentralized control approach. Rigid and semi-rigid formation approaches are presented here. Firstly, a rigid formation is considered, which means that the leaderfollower structure can not be deformed even during obstacle avoidance. In this case, the leader robot is responsible for the safe navigation of the follower robot. In the sequel, a semi-rigid formation is discussed, where the desired leader-follower distance can not be changed during obstacle avoidance, but the desired formation angle can be changed to make the maneuvers possible or even easier. Then, a control strategy based on a fictitious force is presented and implemented on the follower robot to allow changing the formation angle accordingly. To validate the proposed controller, regarding both formation approaches, simulation and experimental results are presented. Applying the described methods in the accomplishment of cooperation tasks, a very good system performance has been observed. Moreover, system stability can be verified through the asymptotic convergence of the formation variables to the desired values during the experiments, which is in accordance with the theoretical analysis performed when designing the controllers.