Research On Robust Consistent Control Of Multi-agent System Based On Neural Network

Multi-Agent System(MAS)is an intelligent distributed system composed of multiple agents interacting in an environment.The main purpose of researching multiagent systems is to expect that multiple agents with relatively simple functions can perform distributed cooperative coordination control and finally complete complex tasks.In the distributed coordination and cooperation control problem of multi-agents,the consensus problem has important practical significance and theoretical value as the basis of cooperative and coordinated control between agents.This paper focuses on two types of problems,the first type of problem is mainly aimed a class of non-linear multi-agent systems at solving the consensus problem under the influence of disturbance factors such as system uncertainty,communication interference and actuator failure.It is assumed that all interference factors are affected by the system's internal and external influences.Internal influences are described in terms of dependence on the state of the system,while external behavior is limited by a constant range.In order to obtain information about the state and constant range adaptive rate,a neural network adaptive mechanism is designed to estimate the rate and range.Based on these estimates,a distributed adaptive sliding mode controller is constructed to eliminate the influence of those interference factors.Then based on Lyapunov stability theory,the consensus of closed-loop adaptive multi-agent system is realized.Finally,the effectiveness of the designed adaptive consensus control strategy was verified by MATLAB / Simulink simulation using a coupled system with four F-18 aircraft longitudinal models.The second type of problem is mainly aimed a class of non-linear leader-following multi-agent systems solving the system by developing a novel neural network learning strategy under the influence of system actuator failure,external additional interference,uncertainty and robust consensus issues.In order to achieve ideal consistency results,the paper proposes a neural network learning algorithm composed of adaptive technology.This algorithm can approximate unknown nonlinear functions and estimate the unknown range of actuator failures.Based on these approximate values and the estimated values,a robust adaptive fault-tolerant consensus control strategy is designed,and then the Lyapunov stability theorem is used to obtain the bounded results of all signals of the closed-loop leader-following system.Finally,the effectiveness of the designed control strategy is verified by MATLAB / Simulink simulation by coupling the nonlinear forced pendulum system.Finally,the fifth chapter of the dissertation summarizes and expects the research of this topic,summarizes the main work done in this dissertation,and proposes the directions for future research.