Stability of discrete sampled systems

Abstract

We consider the wide class of real-time systems that periodically sample their inputs. A desirable property of such systems is that their outputs should be, in some sense, more precise when the sampling period gets shorter. An approximation of this property consists in requiring that, whenever the inputs don’t change, the outputs stabilize after a finite number of steps. We present a set of heuristics to check this stability property, in the case of purely Boolean systems. These heuristics have been experimented on a nuclear plant control software, and have been shown to dramatically reduce the cost of stability analysis.