Fully-Distributed Resilient Consensus of Multi-Agent Systems With State and Input Sector Nonlinearities Under Stochastic Denial-of-Service Attacks

摘要

This paper addresses a fully-distributed consensus control scheme for multi-agent systems (MASs) with state and input nonlinearities under denial-of-service (DoS) attacks. When agents interact with their neighbors, an adversary with malicious intent randomly launches DoS attacks on specific communication channels in the network, resulting in Markovian switching communication topology. Under stochastic DoS attacks, a fully-distributed secure consensus algorithm is proposed using Lyapunov function analysis that is based solely on immediate partial information per node rather than global information. In addition, the developed approach does not require any specific probability or topology pattern under DoS attack. Furthermore, a novel treatment of the input nonlinearity has been established. For this purpose, the input nonlinearity has been transformed, and a condition in terms of a cumulative nonlinearity bound has been formulated via graph characteristics and algebraic properties. Resilient consensus control design conditions have been established by taking into account the input and state nonlinearities. The developed methods only require a spanning tree between MASs for a union of graph topologies under malicious attacks. Moreover, the resultant consensus protocol is fully-distributed by means of an adaptation mechanism, which has been specifically developed for the directed topologies. The efficiency of the consensus approach has been demonstrated using a simulation of robotic systems along with comparison with a closely-related existing work. Note to Practitioners—The purpose of the present study is to provide a control approach for the consensus of nonlinear MASs under DoS cyber-attacks. Nonlinearities in the state equation and control input are considered, which arise owing to the modeling process of physical systems and actuator’s nonlinear characteristics, respectively. A property for dealing with input nonlinearities in MASs has been provided from the consensus control perspective. DoS attacks can be launched by malicious attackers in cyber-physical systems by blocking the communication channels for information flow. A control scheme has been provided by application of time-varying and adaptive parameters to deal with the resilient consensus control of nonlinear MASs under DoS attacks. In addition, a simple design method is established for practitioners to determine the controller gains using numerical routines. The developed distributed consensus method does not require the knowledge of communication topology to find the parameters of the control systems, and such a method is referred to as a fully-distributed consensus approach. The resultant scheme considers the directed graph topologies which do not require bidirectional communication between any two systems. The study has been tested for the consensus of robotic agents under DoS attacks, and the results are compared with the methods available in the literature.