Research On Security Consensus Of Multi-Agent Systems Under Cyber Attacks

Multi-agent systems(MASs)are the new distributed computing systems composed of multiple interactive agents.Each agent is distributed in logical or physical location and usually used for achieving cooperation and completing difficult missions when permitting a set of autonomous agents to work collaboratively with their neighbors.It has greatly improved the ability and efficiency of task execution and has been used more and more extensively in engineering,social economy,computer science,and other fields.Due to the open and shared network communication environment and complex information physical coupling interaction process,the network information risk has become an important factor affecting the stable operation of MASs.Therefore,the research on the security consensus of MASs have important theoretical value and practical significance.This paper mainly studies the security consensus and attack detection of MASs under different kinds of cyber attacks.A novel event-driven mechanism with state-dependent threshold over a given finite horizon is used to update the control input signal.Then a distributed controller is designed to ensure that the MASs can still achieve security consensus in case of denial-of-service(Do S)attacks.And a distributed attack detection filter is designed to accurately detect false data injection attacks(FDIA)in MASs and ensure the stability of the system.The primary research contents of this dissertation are summarized as follows:Firstly,the security consensus problems for time-varying MASs with determined parameters and uncertain parameters based on event-triggered control method under Do S attacks are studied.During information transmission,Do S attacks exist in both control and measurement channels,regarded as a dwell-time switching scheme.To increase the utilization of communication resources,the control input signal can be updated by adopting a novel event-driven mechanism with state-dependent threshold over a given finite horizon.By utilizing multiple Lyapunov functions analysis strategy,sufficient conditions are obtained to ensure the desired consensus performance,and the desired controller is designed via resolving the corresponding linear matrix inequalities.Finally,the effectiveness of this proposed method is verified by providing a numerical simulation.Then,the process of detection of FDIA in time-varying MASs with packet dropout is studied.The / performance index is used to design the FDIA detection module.Further,the network unreliability that comes from packet dropout is modeled as Bernoulli process.Based on stochastic stability,sufficient conditions are obtained to ensure the desired / performance index via resolving the corresponding linear matrix inequalities,and the feasible detector gain can be obtained by solving the multi-objective optimization problem.Finally,the numerical simulation and the actual object model of electric vehicles verify that the attack detector designed in this paper not only has good detection effect but also has strong anti-interference ability.Finally,the main work of this paper is summarized,and the existing problems and future development trend of MASs security consensus research are pointed out.At the same time,future research work is prospected.