Distributed resilient coordination for networked multi-agent systems
Date of Issue2017-06-06
School of Electrical and Electronic Engineering
Critical infrastructures such as smart grid, wireless networks and multi-robot systems, are typical examples of networked multi-agent systems that consist of largescale physical processes monitored and controlled over a set of communication networks and computers. The use of modern information and communication technologies typically adds efficiency, flexibility and scalability. However, it also increases the vulnerabilities to human operator mistakes, physical failures and cyber-attacks. Thus, an urgent need, for safety critical infrastructures, is to enhance resilience against physical failures as well as cyber-attacks. Control systems play a key role in critical infrastructure protection and performance recovery. It is of theoretical significance and practical relevance to develop resilient control algorithms for systems to provide high levels of robustness, reliability and security. This dissertation investigates distributed resilient coordination in networked multiagent systems. It aims to develop algorithms and techniques for networked agents to autonomously coordinate their actions/decisions and achieve a system-level group objective. The background, motivation and preliminaries for this topic are presented in the first two chapters. The problems studied in dissertation include: 1) distributed robust cooperative control; 2) distributed secure coordination. Part I: distributed robust cooperative control. This dissertation starts with fundamental problems of consensus and consensus tracking. In the presence of uncertainties and disturbances, how to develop control algorithms for multiple unknown Euler-Lagrange systems is studied to achieve robust coordination. In such a problem, the agents are not only influenced by interactions among neighboring agents, but also by their own dynamics with unknown uncertainties and disturbances. Identifierbased, continuous, distributed robust control algorithms are developed such that the states of all the agents reach consensus and track a desired time-varying trajectory under different communication topologies, respectively. Next, we address a robust connectivity preserving rendezvous problem for mobile multi-robot systems under unknown dynamics and disturbances. By virtue of a potential field approach, a gradient-based distributed robust control scheme is proposed such that connectivity preserving rendezvous is achieved in the presence of unknown dynamics and disturbances. It is proven that although the multi-robot network has a time-varying network topology, the developed distributed controller is not only able to maintain the connectivity of an initially connected network, but also to guarantee that all the robots can track a desired time-varying trajectory. Part II: distributed secure coordination. This part focuses on distributed control of networked multi-agent systems subject to different types of cyber-attacks. Works on resilience architecture, attack modeling and approaches for specific issues are studied. We firstly consider a distributed secure control problem for systems under two types of switching attacks. Attacks on edges of agents lead to loss of consensus performance. Two distributed secure control laws are developed for solving the issues from a switching perspective. Conditions on secure consensus are derived. Then, this work is extended to secure coordination in the presence of strategic attacks. We model these attacks with their dynamics captured by a random Markov process. A hybrid stochastic secure control scheme with a connectivity restoration mechanism is provided. It is shown that with the proposed design, secure consensus tracking is achieved in a mean-square sense. The effect of strategic attacks on discrete-time systems is also investigated. In addition, we also consider an event-based resilient coordination for linear multiagent systems under Denial-of-Service (DoS) attacks. The DoS attacks refer to interruptions of communications on the control channels carried out by an intelligent adversary. We model a time-sequence-based DoS attack allowed to occur aperiodically in an unknown attack strategy. An explicit analysis of frequency and duration of DoS attacks is investigated for secure leaderless and leader-following consensus issues. An event-based distributed control scheme is proposed and suitable scheduling of controller updating times is determined in the presence of DoS attacks. It is proven that under the proposed controller, the agent group can achieve secure consensus exponentially. In summary, this dissertation investigates resilient coordination techniques for networked multi-agent systems in complex environments. On one hand, when the systems are subject to unknown uncertainties and disturbances, distributed robust cooperative control is studied. To solve this issue, an identifier-based, continuous distributed robust control approach is proposed. On the other hand, when the systems are subject to different types of cyber-attacks, distributed secure coordination is studied. A hybrid distributed secure control scheme is developed. To demonstrate the theoretical significance and practical relevance of our proposed algorithms in this dissertation, distributed multi-robot coordination and distributed voltage regulation of microgrid are investigated, respectively. Finally, brief conclusions and future work are presented in Chapter 8.
DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation::Control engineering