Fault Detection For Multi-agent Systems Under Network Environment

Multi-agent systems have attracted considerable attentions and become a hotspot of control filed due to their development prospects and broad applications.However,in reality,multi-agent systems are vulnerable to faults and attacks from the network because of the typical distributed structure and lack of a centralized entity to monitor the states of the entire system.The fault or the attack occurring at an agent might spread to other agents through the special communication topology of the multi-agent systems,i.e.,a faulty agent might jeopardize the performance of the entire system.Therefor,the security issue of multi-agent systems has aroused many researchers'strong interest.This dissertation systematically investigates the problem of fault detection and isolation for multi-agent systems.The main contents of this dissertation are summarized as follows:1.The problem of fault detection and isolation for discrete-time multi-agent systems without time delay and with time delay are considered by applying disturbance observer theory,respectively.A disturbance observer is designed for each agent,using local information of the agent,to estimate the external adversaries.By utilizing the disturbance observer,the detection of the faults caused by external incipient adversaries can be handled.Numerical simulations are employed to verify the effectiveness of the theoretically analyzed results.2.We investigate the problem of fault detection and isolation for multi-agent systems with disturbances.Firstly,based on unknown input observer,fault detection and isolation for heterogeneous discrete-time multi-agent systems with disturbances is studied.Via the technology of linear matrix inequalities,a sufficient condition is given to ensure the estimation-error residual of the full-order observer using local information is robust to the disturbances.Second,the problem of distributed fault detection and isolation is investigated for multi-agent systems with communication delays and disturbances.A sufficient condition in terms of linear matrix inequalities is derived to make all the estimation-error residuals of the observer,which is designed for multi-agent systems with communication delays and disturbances,satisfy desired H_?performance for communication delays and disturbances.The results show that unknown input observer-based fault detection and isolation investigated in this dissertation can carry out fault detect for signal fault in multi-agent systems with disturbances and time delays.3.The problem of fault detection and isolation is investigated for leader-follower multi-agent systems with disturbances.For the purpose of fault detection and isolation,a sliding-mode observer is firstly designed by using local information and some suitable auxiliary information vector received from neighbor agents.By means of the observer,each agent can estimate the overall state of follower agents even they are not directly connected.Then,using the relative output estimation,a residual vector is developed to detect and isolate the fault occurring on any follower agent of the leader-follower multi-agent systems.By the end,numerical simulations are employed to verify the effectiveness of the theoretical results.