Multi-robot Cooperation Based-on Personality

With flexible architecture and high adaptability, multi-robot system (MRS) can be widely used in many areas such as astronautics, national defense, industry, service and so on. To take the advantages of MRS, system coordination is a key problem. The dynamic and complex characteristic of the multi-agent system (MAS) always leads to dimension explosion problem, so the coordination effect will be poor if we only directly generalize the traditional negotiation, learning or evolution methods in single agent system to MAS. How to find the intrinsic characteristics of MRS and get a new way to coordinate the system is an important subject in current MRS research.In. this dissertation we believe that the diversity is an important but often neglected characteristic of MAS. Through the analysis of diversity, the concept of 'personality' in psychology is introduced into the MRS. The intention, behavior inclination and different degree of ability of different robots are described systemically, and a new cooperation method based on the personality is put forward. Focusing on the idea, the main contributions of this dissertation are:1. A survey is given on current research areas and status of MRS. The behavior implementation based on the behaviorism is introduced emphatically.2. Diversity in nature and social system is analyzed and an evaluation criterion of diversity is given. It is pointed out that diversity is not only a ubiquitous characteristic of the MAS, but also very important in enhancing adaptability of the system.3. The diversity and efficiency of the homogeneous forage, color forage and territorial forage in MRS is analyzed. Simulation results show that the diversity will not certainly cause the improvement of the system performance, if we want to make use of it, the key factors related with the system performance should be considered considerately.4. Personality, a concept in psychology, is introduced into the MRS to describe one of the important aspects of diversity in MRS. The mathematical description and evolution method of the robot personality is given. And the effect of personality onthe robot behavior mode and system cooperation is discussed.5. A self-organizing conflict avoidance method using the difference resulted from the robots' personalities is proposed. The application of this method in competition environments, such as forage and movement planning, are studied by simulations.6. A new tracking method P-CMOMMT based on altruism for Cooperative Multi-Robot Observation of Multiple Moving Targets problem is given. The effect of altruism on the tradeoff between the individual benefit and the collective benefit is studied. Simulations show the P-CMOMMT exhibits better tracking performance than the traditional CMOMMT and A-CMOMMT methods.7. The performance of the centre reference point method and leader reference point method in multi-robot formation is studied. And the influence of the altruism degree on the velocity selection of leader reference point method is analyzed.8. A new self-organizing task allocation algorithm PJTAM is proposed. It is based on the Ant Colony Optimization (ACO) and has a persistence factor. Simulations show that P_TAM solves the task deadlock problem successfully and is quicker and more robust than the classical methods based on threshold.Simulation results in the typical research areas above show that, the cooperation algorithms based on personality put forward in this dissertation have high coherence and adaptability, and are easy realization. It provides a new way to solve the classical problems such as conflict avoidance, self-organizing task allocation and tradeoff between the individual benefit and the collective benefit of the multi-robot system.