Distributed secure state estimation for linear systems against malicious agents through sorting and filtering

Abstract

This paper investigates the distributed secure state estimation problem for the cyber–physical systems monitored by a multi-agent network where partial agents are malicious. Through introducing a sort and filter approach, a novel distributed secure state estimation strategy, where the impact of malicious agents is mitigated through discarding partial extreme values from the received vectors, is proposed. Sufficient conditions on the graph topology and the system matrices to tolerant a bounded number of malicious agents are given. It is shown that by adopting the proposed secure state estimation strategy, normal agents can effectively generate correct state estimate in the presence of malicious agents. Besides, for efficiently updating the state estimate while new measurements are available, a distributed observer is also designed. Compared with the existing techniques searching an appropriate candidate from multiple ones, the proposed secure state estimation strategies are much more computationally efficient since reliable state estimates are generated without brute force search. Finally, simulation results are provided to illustrate the effectiveness of the proposed algorithms.