K-loss robust codiagnosability of Discrete-Event Systems

Abstract

Recently, the problem of robust diagnosis against loss of event observations has been proposed in the literature, where it is assumed that some sensors, or communication channels between sensors and diagnosers, are subject to failures. In these works, permanent or intermittent failures are considered, and models of the plant subject to these failures are obtained. One characteristic of the robust diagnosis method considering intermittent failures is that the faulty sensor may or may not recover from the failure, and the intermittent failures may evolve to permanent failures, which are also represented in the model of the plant subject to intermittent failures. However, in some cases, the communication failure is transient due to external interference, i.e., the communication channel always recovers from the failure after a short period of time. In this paper, we formulate the problem of robust diagnosis against transient sensor communication failures. The new formulation leads to the definition of K-loss robust codiagnosability. We also present a method to verify this property.