On-line Detection and Sensor Activation for Discrete Event Systems1

Abstract

This paper investigates on-line detection and sensor activation problem of discrete event systems. The objective is to derive minimal sensor activation policy while preserving strong detectability and strong periodic detectability. After reviewing strong detectability and strong periodic detectability, we introduce a new concept called distinguishability. Distinguishability is used for two purposes. First, it is used to characterize detectability. Secondly, it is used for on-line state estimation and sensor activation. To this end, we proposed an on-line state estimation framework which estimates the state of the system. Sensor activation is then achieved based on the state estimation. State-estimation-based sensor activation has some interesting and surprising properties that we will investigate in this paper. In particular, we introduce a new concept called coherence that plays a key role in state-estimation-based sensor activation. To obtain state-estimation-based sensor activation policy that is minimal and coherent, two algorithms are derived. One algorithm is for strong detectability. It minimizes sensor activation while preserving distinguishability, and hence strong detectability. The other algorithm deals with strong periodic detectability. For strong periodic detectability, a new property called information-preserving must be incorporated. Both algorithms are of polynomial complexity.