Multi-robot Cooperation Based On Biological Immune System

With the limitation of information acquisition, processing and control for a single robot system's capability, multi-robot system (MRS) is proposed to perform the tasks, which are impossible for a single robot, and increase the working efficiency. Taking advantage of the MRS, such as the distribution, adaptability, robustness and high efficiency, carrying out complicated tasks in uncertain and dynamic environments has become a main research direction in robotics. MRS can be widely used in many fields, including industry, aerospace, military affairs, social service and so on. However, the multi-robot system is not just the simple gathering of several robots, which will decrease the working efficiency of the MAS instead of making use of the above advantages of the MAS. Therefore, how to efficiently organize the multi-robot group and make full use of the cooperation of the MAS is an important subject in current MRS research.The initial invention of the robot is to imitate the intelligence of the human being and other species, to help and substitute people to work. Therefore, to realize better cooperation among robots, the MRS is expected to get inspirations from human society and biological colony. Based on the recognition, learning, memory capability and the advantages of distribution, dynamic-balancing, adaptability of the immune system, we utilize the working mechanism and characteristics of the biological immune system, and study the important areas of the multi-robot system, including the architecture, cooperation, task allocation and learning areas. MRS improves the cooperation of robots and the adaptability for the dynamic environment based on the immune system. The main contributions of this dissertation are:1) A summary is given on the state-of-the-art technologies on MRS. Some typical application areas are introduced related to the cooperation of MRS, especially the biological research of MRS.2) The characteristics and functions of biological immune system_are introduced. The research and application of the artificial immune system on robotics are summarized. And based on the principles of the immune system, the multi-robot architecture and single-robot architecture of the MRS are designed as the basic framework of this dissertation. The cooperation of multi-robots is proposed and analyzed based on the immune architecture.3) Focus on the decision-making module of the architecture. Based on the information fusion of the ultrasonic sensors, the immune based exploration algorithm for the MRS is proposed to realize the complete exploration and map building in the uncertain environment. The algorithm not only realizes the complete exploration, but also efficiently avoids the conflict and interference of the robots, and decrease the repeat coverage rate. In the simulation, the efficiency of the algorithm and the robustness of the system are validated.4) To accomplish the unknown tasks, a new multi-robot cooperation algorithm is designed in an uncertain environment based on the self-reinforcement adjustment of antigen stimulus and interactions of antibodies. During the simulation, robots push the boxes randomly distributed, which are unknown in advance, to the destination by autonomous cooperation. And the deadlock problem is solved in the MRS. The study is also belong to the decision-making module of the architecture.5) Focus on the task-allocation module of the architecture. To overcome the greedy deficiency of the current task allocation algorithms of the MRS, an immune based task allocation algorithm is proposed. The algorithm fully utilizes the antigen(task) information, the interaction among the antibodies from the same robot and different robots. The experiments of autonomous emergency handling validated the efficiency of the task allocation algorithm.6) Focus on the learning module of the architecture. To increase the intelligence of the robot system to adapt the dynamic environment, by utilizing the self-learning, self-adaptability and memory ability of the immune system, the immunized reinforcement learning algorithm is proposed and applied to the robot system. Furthermore, based on the immunized learning of a single robot system, cooperation among multi-robot system is considered and applied to Cooperative Multi-Robot Observation of Multiple Moving Targets(CMOMMT) problem. Simulations validated the learning capability and the dynamic cooperation of the multi-robot system based on the learning mechanism.Research on the typical areas above shows that, multi-robot cooperation based on immune system has autonomy, cooperation, adaptability and robustness. It provides a new way to solve the typical problem in multi-robot system, such as cooperation, task allocation and learning.