Secure distributed control in unreliable D-NCS

Abstract

Distributed Networked Control Systems (D-NCS), such as the electric power system, the transportation system, or almost any large-scale network, are vulnerable to cyber attacks. The compromised nodes in the D-NCS can affect it and cause the distributed control algorithms (e.g., consensus algorithm) to fail. This paper is concerned with the problem of designing a secure distributed control methodology that is capable of performing a secure consensus computation in a D-NCS in the presence of misbehaving nodes. We consider the case of formation control in a multi-robot system using the linear consensus algorithm, and we model the malicious attack as an exogenous input that compromises the behavior of a single robot in this multi-robot system. The proposed secure distributed control methodology includes four phases: (1) Detect the neighbors' misbehaviors relying only on each robot's local observations; (2) Adjust the consensus computation weights according to the neighbors' reputation values; (3) Identify and isolate the compromised robot; and (4) Update the reference state using the adjusted consensus computation weights to ensure the convergence of well-behaving robots. A Simulink-based testbed for multi-robot formation control is used to illustrate the effectiveness of the proposed method.