Networked multi-agent consensus under imperfect communication and node failure

Abstract

Recently, multi-agent networks, such as distributed robots and mobile sensor networks, have been widely studied due to their broad application. The main focus of studying multi-agent networks is on how consensus has emerged as a result of local interactions among agents, where consensus means a collection of agents reaching an agreement on certain quantities of interest. Most of the research literatures on consensus of multi-agent networks mainly concentrate on the analysis of network models with perfect communication, which assume that each agent can receive information from its neighbours timely and accurately. However, such models do not reflect most real circumstances, as the information flow between two neighboring nodes is generally affected by many uncertain factors including limited communication capacity, network induced time delay, random packet loss, and so on. Moreover, in many realistic networked multi-agent systems, due to the complexity of systems and external attacks (disturbance), the failure inevitably occurs in nodes or links. The aforementioned communication constraint should be considered in the design of control strategy or algorithms. In this thesis, we investigate the networked multi-agent consensus problems under imperfect communication and node failure. The following issues will be presented in this thesis in detail: (a) continuous-time multi-agent network consensus with quantization and time delay; (b) discrete-time multi-agent network consensus with quantization and time delay; (c) distributed event-triggered control approach for discrete-time multi-agent networks; (d) consensus recovery approach to nonlinear multi-agent system under node failure. The main contributions of this thesis are listed as follows: ● A precise mathematical treatment for the continuous-time multi-agent network with quantization and time delay is presented. Based on the nonsmooth analytical technique, the existence of the global solution to the resulting system is strictly proved in the Filippov sense. Moreover, the effect of the time delay and quantization on the consensus results is studied and the explicit relationship among time delay, quantization parameter and the practical consensus set is explicitly presented. ● A new discrete-time consensus protocol considering the effect of quantization and time delay simultaneously is proposed. Consensus of discrete-time multiagent network under two different quantizers are respectively investigated. Some interesting consensus results are derived for the different quantizers. ● Event-triggered control strategy was proposed to deal with the discrete-time multi-agent consensus problem. For the network with or without time delay cases, the centralized or distributed event-triggered conditions are proposed, which can ensure the consensus of the proposed multi-agent model. ● In large-scale networks, node failure is unavoidable and generally has undesirable effects on the stability and performance of the systems. To improve the reliability of the network system under node failure, a new consensus recovery approach will be provided. After the consensus recovery operation, the consensus property is well reserved. Theoretical consensus analyses have also been presented by fully utilizing the network structure which can verify the correctness of the consensus recovery approach.

Date of Award	2 Oct 2013
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Wing Cheong Daniel HO (Supervisor)