Iterative Learning Control Of Multi-agent Consensus

Multi-agent system consists of a set of intelligent agents that mutually communicate and cooperate to complete a task. In which, every agent is a physical or abstract entity, and acts on its own and environment. Compared with single agent system, it has stronger distribution and autonomy. Its distributed coordination control has attracted considerable interests due to its wide applications in many areas such as autonomous underwater robots, mobile robots, unmanned air vehicles, distributed sensor array and so on.For repetitive and distributed multi-agent system, this paper uses iterative learning control method to achieve perfect tracking in a finite time interval. The Innovations of this paper is to propose two iterative learning control algorithms for initial state problem in iterative learning control. The convergence of the two algorithms is proved through theoretical analysis. And Matlab simulation also proves the reliability of two algorithms. Finally, for the experiment platform of three rigid robot manipulators, we use two methods to achieve formation of the end-tip of the robot manipulators. Good experimental results prove the validity of the proposed algorithms again. Detailed content of this paper is as follow.For a distributed multi-agent system which performs a given repetitive task, only a portion of the agents have access to the desired reference trajectory. And the initial state of every agent is arbitrarily fixed. A PD type distributed iterative learning control algorithm combined with initial state error correction of every agent is proposed to solve this consensus problem within a finite time interval. This initial state error correction acts within only a period of timet∈[0, h], and finally every agent in system achieves perfect tracking over a finite time interval t∈[h, T].The convergence of this algorithm is proved through theoretical analysis. Finally simulation results are presented to illustrate the performance and effectiveness of the iterative learning algorithm.For a distributed multi-agent system executing a given repetitive task, only a portion of the agents have access to the desired reference trajectory. On the condition of the unknown initial states corresponding to the desired trajectory, a distributed learning control algorithm combined with initial state correction of every agent is proposed for the system with interference. The convergence analysis shows that the proposed algorithm can eliminate the tracking errors caused by different initial states between every agent and the desired trajectory, then perfect tracking control within the finite-time interval is available. The simulation results also prove the effectiveness of the proposed algorithm.Experiments are carried out on the platform of three rigid robot manipulators. One kinematic model for each manipulator is set up by chosing two appropriate joints. The two proposed algorithms are implemented on the platform based on the established models. And experimental results show that reference trajectory of the end-tip is reproduced consistently for each manipulator.