Research On Secure Consensus For Multi-agent Systems Under Malicious Attacks

In recent years, multi-agent systems have been widely applied in a large range of engineering fields such as wireless sensor networks(WSNs), intelligent transportation, robot cooperative control, unmanned aerial vehicles(UAVs) formation flight, etc. Multi-agent systems are penetrating into our daily lives, changing the way we work and live. The distributed coordinated control of multi-agent systems, as one of the key technologies of these applications, has attracted widespread concerns and has became one of the hottest research topics.Consensus is a representative problem in the distributed coordinated control of multiagent systems, and is also the basis of many other decentralized control and estimation problems, whose goal is to make the states of all agents asymptotically achieve an agreement as time goes to infinity according to exchange the local information with their neighboring agents. Note that in complex network environments, multi-agent systems may easily suffer from some malicious attacks. Therefore, as one of new research fields in multi-agent systems, the security issues of consensus problems has aroused many researchers’ strong interest.Based on the study of previous works, this dissertation systematically and deeply investigates the security issues of consensus control for multi-agent systems under malicious attacks, by applying control theory, stability theory, matrix theory as well as graph algebra theory.The main contents of this dissertation are summarized as follows:1. An overview of existing literature on secure consensus problem of multi-agent systems is provided, and some deficiencies of this research field have been analyzed.2. The secure consensus problem for discrete-time multi-agent systems under communication delays is investigated. For both normal and malicious agents in the network,a novel graph-theoretic property referred to as topology robustness is introduced. Then,based on the topology robustness and malicious attack model, a distributed adaptive secure consensus protocol is given. For the considered networked system, the control input of each normal agent can only use its own value and the delayed information of its neighbors. Sufficient and necessary conditions, which depend on the control parameters, the topological property, and the communication delay, are obtained to guarantee the final convergence of the secure consensus protocol.3. The secure consensus problem for nonlinear continuous-time multi-agent systems suffering from attacks and communication delays is investigated. Consider a directed multiagent network composed of n agents with nsnormal agents and na= n- nsmalicious agents. A novel delay robust secure consensus algorithm according to the neighboring nodes’ delayed information is designed. Convergence analysis of the system under the protocol designed is provided by using Lyapunov-Krasovskii stability theory and Barbalatlike argument approach.4. The finite-time secure consensus problem for multi-agent systems under attacks is investigated. Based on the maximum number of malicious agents in each normal agent’s neighborhood and topology robustness, an effective method for choosing neighboring agent’s information is designed. Then, by combining the iterative learning control approaches,a finite-time secure consensus protocol is proposed. The theorem result shows that the proposed protocol can guarantee all normal agents to resist the malicious attacks and achieve the finite-time consensus if the network topology has sufficient connectivity in terms of robustness.5. The secure consensus problem for multi-agent systems with quantized communication is investigated. A novel quantized-data based secure control law with built-in security mechanism is proposed to achieve consensus in the presence of attack agents. Sufficient conditions for secure consensus protocol under f-locally bounded attack model have been provided.