Statistical Model Checking of Incomplete Stochastic Systems

Abstract

We study incomplete stochastic systems that are missing some parts of their design, or are lacking information about some components. It is interesting to get early analysis results of the requirements of these systems, in order to adequately refine their design. In previous works, models for incomplete systems are analysed using model checking techniques for three-valued temporal logics. In this paper, we propose statistical model checking algorithms for these logics. We illustrate our approach on a case-study of a network system that is refined after the analysis of early designs.

Verification of Variability-Intensive Stochastic Systems with Statistical Model Checking

“…Given a finite execution, the approach can either decide if it satisfies the property by comparing the outcomes of two finite automata, or return an undefined value in case the comparison is inconclusive. Such a three-valued approach cannot be handled by classical Monte Carlo algorithms, but promising extensions exist [1] and can inspire our work.…”

“…In the end, the only difference in our working assumptions -which has no incidence on the proof -is that the number of paths that each variant can execute can differ. This is taken into account with the profile 1 n and is equivalent to having paths that some variants executes with a zero probability. For the sake of conciseness, we do not replicate the proof of Delahaye et al here.…”