This paper proposes a path planner for multi-robot systems based on the solution of the multiple traveling salesman problem by genetic algorithm. The main planning goal is to build an efficient path to each robot of the system which jointly ensures that several a priori known points (ways-points) in the environment can be attained by at least one of the robots. During navigation, each robot try to follow its planned path while performing tasks and deviating from unknown static and dynamic obstacles. As the robots have limited sensing and communication skills, their positions can easily become uncertain and the robots will be unable to follow their planned paths. To circumvent this problem, each robot has a local planner, able to recalculate its position and then to resume its planned path. This computation is based on the knowledge about way-points positions, information exchanged with other robots and a self localization algorithm by triangulation. The proposed global/local path planner is firstly validated by computer simulations. An experimental study is also carried out with a small system with very simple mobile robots with differential drive and Bluetooth communication. The obtained results confirm the efficacy of the proposed path planner.

本文提出了一种基于遗传算法的多旅行商问题求解的多机器人系统路径规划器。主要的计划目标是为系统的每个机器人建立一条有效的路径，以共同确保至少一个机器人可以在环境中获得多个先验已知点（路标）。在导航过程中，每个机器人在执行任务并偏离未知的静态和动态障碍物时都会尝试遵循其计划的路径。由于机器人的感测和通信能力有限，因此它们的位置很容易变得不确定，并且机器人将无法遵循其计划的路径。为避免此问题，每个机器人都有一个本地计划器，可以重新计算其位置，然后恢复其计划路径。该计算基于以下知识：航路点位置，与其他机器人交换的信息以及通过三角测量的自定位算法。首先通过计算机仿真对提出的全局/局部路径规划器进行了验证。还对一个小型系统进行了实验研究，该小型系统具有带差动驱动和蓝牙通信的非常简单的移动机器人。获得的结果证实了所提出的路径规划器的功效。