Research On Motion Control Of Multiple Mobile Robots System

With the development of robotics, the capabilities of robot have been improved ceaselessly and its application areas have been extended greatly. Robots are being expected to do more complicated tasks, such as exploration under deep ocean and even the treatment of nuclear industry fault, operation in outer space etc. On the one hand, when the tasks are too complex to be set up with a single robot, they can be accomplished through the coordination and collaboration of multiple robots. On the other hand, the coordination and collaboration among multiple robots can improve the efficiency of robots system and make the system be more adaptive and fault tolerant. The coordinate control of multiple mobile robots is always a hot and essential topic in multi-robot system domain. In this dissertation, the architecture, path planning and some typical coordinate tasks of multi-robot system, such as formation control and pursuit-evasion game, have been discussed thoroughly. The main contents and contributions of this dissertation are as the followings:Based on the survey of the research status at home and abroad, one contribution is to establish the architecture for multi-robot system, which defines the relationship and function allocation among multiple robots in the system, specifies the information flow and the logical topological structure between the system and the robots, and presents the mechanism and algorithm structure which can control multiple robots and make them act cooperatively. According to the characteristics of Pioneer 2-Dxe intelligent mobile robot, a multi-robot coordination control system based on the open agent architecture (OAA) is presented, and a physical multi-robot experiment platform is constructed, using several Pioneer 2-Dxe mobile robots as hardware platform and OAA2.3.1 as software platform. This system has some advantages, such as openness, plug and play, distributed computation and multimodal, which can satisfy the requirement for multi-robot system working in unknown complex environment.A novel global path planning method for mobile robot based on particle swarm optimization is proposed in the dissertation. First the free motion link graph is built for the working space of the mobile robot by using free space method, and the shortest path from the start point to the goal point in the graph is obtained by Dijkstra algorithm. Then PSO is adapted to optimal the path that already got. Aiming at the shortcoming of the PSO, which is, easily plunging into the local minimum, the dissertation puts forward an advanced PSO algorithm with the mutation operator.Multi-robot path planning has to deal with not only the obstacle avoidance problem, but also the collision avoidance problem. According to multi-agent system theroy, a distributed multi-robot path planning algorithm based on gated dipole model is presented. Each robot as an agent uses the modal to plan its optimal path independently and obtain the environment information by sensor and the communication between other robots. The robots use dynamic priority method to avoid the collision with each other.Formation control is a typical multi-robot coordinational problem. A behavior based arbitrary formation control algorithm is discussed. The behavior based method exist only local formation feekback, and cannot guarantee the stabilization of the formation. Considering the OAA character, the dissertation proposed a formation control strategy with global formation feedback and it is proved that this control strategy is asymptotical stable by Liapunov stability theory.Reinforcement learning is a novel mechine learning method and has been widely used in robot domain. To carry out a coordination task that multiple robots surround multiple invaders, a multi-agent individual reinforcement learning algorithm is used. An experiment simulation platform consisting of many robots and targets is built. Several experimental conditions are designed and their influences on the robots'group behavior performance are discussed. PSO algorithm is used to find out the optimal experiment configuration under different optimal criterions.