Research On Multi-Agent Systems With Combined Privacy Preservation And Resilience Consensus

The characteristics of autonomy,intelligence,and collaboration possessed by multi-agent systems(MASs)have made them one of the important forms of distributed artificial intelligence applications.Security consensus is a key cooperative control problems in MASs.Existing security consensus control methods typically discuss scenarios in which there are only privacy eavesdroppers or malicious attackers in the system.However,MASs not only has the risk of privacy disclosure but also is exposed to malicious cyber attacks in practical applications.This thesis focuses on the unresolved issue of effectively combining privacy preservation and resilient consensus in MASs.Using methods such as graph theory,state decomposition and distributed cooperative control,this thesis conduct research on MASs with combined privacy preservation and resilient consensus.A joint design control method is proposed for node initial state privacy protection and resilience average consensus in undirected graph MASs under deception attacks.The method first proposes a distributed node initial state value reconstruction mechanism based on state decomposition to protect the privacy of node initial state information;Secondly,combining the reconstructed node state and Secure Acceptance and Broadcasting Algorithm,a deception-resistant MASs resilience average consensus control method is proposed;Integrating the aforementioned reconstruction mechanism and anti-deception attacks method to achieve joint design for consensus control.Theoretical analysis and simulations have verified the effectiveness of the proposed method in protecting the initial state privacy of nodes and achieving resilient average consensus.A joint design control method is proposed for node initial state privacy protection and resilience average consensus in directed graph MASs under multiple cyber attacks.The method first proposes a fully distributed improved state decomposition mechanism to protect the privacy of node initial state information;Secondly,according to the characteristics of different types of cyber attacks,an adaptive MASs resilience average consensus control method is proposed to resist multiple cyber attacks;Integrating the aforementioned improved state decomposition mechanism and anti-cyber attacks method to achieve joint design for consensus control.Theoretical analysis,simulations,and Raspberry Pi application experiments have verified the effectiveness of the proposed method in protecting the initial state privacy of nodes and achieving resilient average consensus.A joint design control method is proposed for node whole process state privacy protection and resilience consensus in undirected graph MASs under Sybil attacks.The method first proposes a fully distributed lightweight encryption mechanism for the whole process of node state information to protect the whole process privacy of node state interactions;Secondly,based on the degree-based mean-subsequence-reduced,a resilience consensus control method for MASs against Sybil attacks is proposed;Integrating the aforementioned encryption mechanism and antiSybil attacks method to achieve joint design for consensus control.Theoretical analysis and simulations have verified the effectiveness of the proposed method in protecting whole process state privacy of nodes and achieving resilient consensus.Finally,the thesis is summarized,the shortcomings of the research are pointed out,and the future research directions are discussed.