Research On Task-partitioned Multi-robot Coordination And Tracking Control

Multi-robot system (MRS) has the advantages of strong fault-tolerant ability and high robustness, it can complete task efficiently and reliably under complex environment, so it has been paid much attention nowadays. Some problems, such as task assignment, resource competition, conflict resolution, and so on, emerge in this system. Only by coordinatively controlling the robots can these problems be solved. By this way the performance of the system will be enhanced notably, the efficiency of the system will be increased greatly, and the function and the operating region of the system will be extended obviously. Because of MRS's simple structure, energy-conservation, easiness to control, and higher practicality, the multi-robot system, which is composed of several wheeled mobile robots having perceptivity and decision-making capability, is the most typical system with good application prospects in coordinative MRS. Combined with the mechanism of immune system and the cooperation method of multi-agent system, this dissertation gives a deep and systematic research on the architecture of coordinative MRS, the multi-robot coordination under oppositional and cooperative environment, and the tracking and control of multi-robot. The main research works are as follows:1. Based on the analysis of the tasks in MRS, the concept of task-partitioned coordination is given, and its corresponding classification is proposed. Through thoroughly analyzing the structure and task characteristic of MRS, a hierarchical distributed coordination model is presented, and a structural framework of the taskpartitioned coordinative multi-robot system is constructed.2. MRS is composed of individual robot. The coordination of multi-robot is closely related to the way by which individual robot is controlled. Based on the research on existing individual control structure, the mixed robot individual control architecture which unifies the behavior control and the cognition structure is proposed by introducing the immunity selection and arbitration mechanism into the robot control architecture. MRS can not only react to the dynamic changes of environment and task, but also have learning and adaptive ability to complete tasks by cooperation. Combined the immune network with the dynamic characteristic of each robot in MRS, the concept of multi-robot's synthesized immune network (SIN) is proposed. Based on the Jerne's immunity network model, the SIN model is constructed, and its algorithm is designed. The simulation experiments indicate that the algorithm is correct and effective.3. The multi-robot's cooperation based on mixed task is studied further. First multi-robot's cooperation between two competing parts with competition task but no communication is studied. After analyzing the operation environment of multi-robots, a MRS competition task coordination algorithm is proposed based on adversarial equilibrium. Then based on the competition and cooperation task algorithms, the coordination method of mixed task multi-robots is proposed. A multi-robot coordination framework and corresponding algorithm based on task-partition are proposed. Zero-sum Markov games are adopted to resolve robots' competition task between unions, and SIN algorithm is adopted to accomplish robots' cooperation task within the union. Experiments on simulation soccer robot platform show that the mixed task coordination algorithm is feasible and superior.4. A compound tracking and control method for robots system is put forward. The method combining an immune feedback mechanism with non-time based control approach not only keeps the merit of non-time based approach, but also can fully consider the requirements of the mechanical structure and the characteristics of motor torque of robot, so it can limit the peak-to-peak value when the error is bigger. By selecting appropriate parameters, the performance of the control system can be improved. For the new dynamic event, the insertion processing algorithm is used to enable the robot to adapt the dynamic changing environment quickly. This dissertation expands this method to solve the multi-robots control problem. The queue control simulation and the experiment results indicate that the expanded method have better tracking ability and smoother control curve. Thus the compound tracking control method is more adaptive to practical application environment.