Distributed Secure Cooperative Control Of Multi-agent Systems Under Cyber-attacks

With the progress of sensing and communication technology,networked multi-agent systems have at-tracted an increasing number of attention from researchers,which have advantages of high control efficiency and strong scalability.However,due to the complexity of network environment,some security risks appear,which leads to that the distributed cooperative control of networked multi-agent systems becomes more com-plicated.Therefore,this dissertation mainly studies the distributed secure cooperative control of multi-agent systems under cyber-attacks.The main research contents include:secure consensus control of discrete-time multi-agent systems against false data injection attacks on actuators and sensors,H_∞control of discrete-time multi-agent systems under stochastic denial-of-service attacks with partially known attack probability,secure output consensus control of continuous-time heterogeneous multi-agent systems against false data injection attacks,observer-based secure output consensus control of discrete-time heterogeneous multi-agent systems against two types of false data injection attacks,secure output containment control of continuous-time het-erogeneous multi-agent systems against cyber and physical attacks.The content of the dissertation can be summarized into the following five aspects:(1)The secure leader-following consensus control problem for discrete-time multi-agent systems against false data injection attacks on both actuators and sensors is studied.It is shown that under standard consensus protocol,an attacked follower can cause normal followers to fail to follow the leader if the normal followers can access the attacked follower through directed paths of arbitrary length.To mitigate the adverse impacts of false data injection attacks,an observer-based secure consensus protocol is proposed where local observ-er is used to estimate the system state and attack signals simultaneously and then the uniformly ultimately boundedness of state consensus error between all followers and leader can be guaranteed.Finally,a simula-tion example validates the unavailability of standard consensus protocol and the effectiveness of the proposed observer-based secure consensus protocol.(2)The H_∞leader-following consensus control problem of multi-agent systems under stochastic deny-of-service attacks whose attack probabilities are not completely known is investigated.When agents commu-nicate with each other,an attacker with partially available attack information stochastically launches deny-of-service attacks into some channels in the communication network,which makes the communication topology Markovian switching with partially unknown transition probabilities.By applying Lyapunov function analy-sis,a sufficient condition is given to ensure that the leader-following consensus can be achieved in stochastic stability sense with a prescribed H_∞performance and the feasible consensus control gain can be figured out.Finally,a simulation example of LC oscillator circuit network is given to illustrate the feasibility of the results.(3)The secure output consensus problem of heterogeneous multi-agent systems against cyber-attacks is researched.Both the sensors and actuators of followers may be injected into false data injection attacks.To mitigate the malicious impacts of these attacks on output consensus performance,a desired reference model which is combined with an auxiliary system is first designed for each follower to simulate its normal system dynamics.Then,based on this reference model,a state feedback cooperative controller which consist of adaptive compensators are proposed.The integrated control protocol can ensure that the output consensus error is arbitrary small by adjusting some designed parameters even in the presence of false data injection attacks.An illustrative example is employed to demonstrate the effectiveness of the proposed method.(4)The secure output consensus problem of discrete-time heterogeneous multi-agent systems against two types of cyber-attacks is studied.The attackers continuously launch attack signals into the actuators and stochastically launch attack signals into the sensors with unknown Markovian attack probabilities.An observer-based cooperative output regulation consensus controller is designed to mitigate the detrimental influence of actuator attack on consensus performance.To avoid corrupted sensor’s signal being used by the observer,an attack detector based observer is developed for each follower to estimate both attack signal and system state.By converting the estimation error system subject to an unknown Markov modeled attack to a Markov switching system with unknown transition probabilities,a sufficient condition of the observer gain is given.Moreover,a discrete-time distributed auxiliary system is designed to estimate the leader state.Finally,a numerical example and a practical example of heterogeneous mobile vehicles show the effectiveness of the proposed control method.(5)The secure output containment problem for multi-agent systems composed of homogeneous leaders and heterogeneous followers against attacks in cyber and physical layers is studied.In the physical actuators of followers,malicious attackers intentionally inject false data injection attacks with unknown time-varying signals to disrupt the cooperation of the systems.To eliminate the adverse effects of the false data injection attacks,an adaptive compensator is designed for each follower.In the network channels among agents,adver-saries launch replay attacks such that the agents can only retrieve the previous information of their neighbors.To characterize the effects of replay attacks,a distributed auxiliary system with heterogeneous time-varying delays is constructed to estimate the convex combination of the leaders.Based on the proposed adaptive compensator and distributed auxiliary system,a secure containment control protocol is proposed to ensure that the output trajectories of each follower can converge to the dynamic convex hull spanned by the leader outputs.Finally,an illustrative example is given to demonstrate the effectiveness of the proposed method.