Research On Immunity-based Multi-robot Cooperative Exploration

With the fast development of science and technology, robots have been applied from manufacturing industry to aerospace & aviation, military and so on. Robots can carry out their intelligent work autonomously in unknown environments. The key problem of map-building and other tasks for autonomous robots is exploration. Compared with single robot, multi-robot system has a lot of advantages. However, exploration strategy of multi-robot system, which should take into account of cooperation of multiple robots, is more difficult than single robot. This dissertation studies cooperative exploration strategy of multi-robot system in unknown environments.Firstly, the dissertation reviews the research on multi-robot exploration and its relative aspects in the world. The background and the main contents of this dissertation are described briefly.Secondly, the key to multi-robot exploration is the distribution of targets for multiple robots. Aiming at the combinatorial optimization problem, Burgard presented a decision-theoretic approach to coordinate the robots. However, the computation burden for optimal allocation increases exponentially with the number of robots and target points. To solve this problem, we propose a genetic algorithm-based coordinated multi-robot exploration algorithm with its characteristics of random global searching to seek the solution quickly. To get over the premature convergence of simple genetic algorithm, we present a multi-robot cooperative exploration strategy based on the improved genetic algorithm. The selection probability is computed based on the similarity vector distance to guarantee the antibody's diversity. The crossover and mutation probability is adjusted based on the fitness of antibody to decrease the possibility of local optimal.Thirdly, in order to get over the drawbacks of centralized structure and improve the scalability of multi-robot systems, the dissertation learns from the distributed architecture and adaptive balance ability of the immune system and establishes dynamic distributed multi-robot system. Combining with the interaction mechanism between antigen and antibody and among antibodies, this paper proposes a distributed multi-robot network exploration strategy based on the immune principle.Fourthly, an improved clonal selection algorithm based on the clonal selection principle is proposed for multi-robot exploration. Because the computation burden for the binary-coded clonal selection algorithm is heavy, the real-coded clonal selection algorithm is adopted. With the help of clone and adaptive Cauchy mutation operator, the speed of convergence and diversity of the antibodies is improved apparently. All of the above algorithms are proved and compared with the extensive simulation experiments.Finally, conclusions are given with recommendation for future work.