Verification of Stochastic Timed Automata

Abstract : Verification is now a well-known branch in computer science. It is crucial when dealing with computer programs in automatic systems: we want to check if a given system is correct and satisfies some specifications that should be met. One way to analyse those systems is to model them mathematically. The question is then: can we check if the model satisfies the required specifications ? This is called the model-checking problem. Several models have been studied in the literature. We have an interest for models that can mix both timing and randomized aspects. In this thesis we thus study the stochastic timed automaton model (STA). The contributions of this document are twofold. First, we study the qualitative and quantitative model-checking problems of STA. STA are, in particular, general probabilistic systems and with such model, one is thus interested in questions like « Is a property satisfied, within a given model, with probability 1 ? » (qualitative) or « Can we compute an approximation of the probability that the model satisfies a given property ? » (quantitative).We study those questions for general stochastic systems using, amongst other, the notion of decisiveness used in infinite Markov chains in order to get strong qualitative and quantitative results, and that we extend here in or more general context. We prove several results for the qualitative and quantitative model-checking problems of those probabilistic systems, some of them being extensions of previous work on Markov chains, others being new, and we show how it can be applied to subclasses of STA. Then we study the compositional verification in STA. In general, a system is the result of several smaller systems working together. Compositional verification allows then one to reduce the analysis of a big system to the analyses of the smaller systems which compose it. It is then crucial to have a good compositional framework in mathematical models, and this lacks in STA. In this thesis, we define an operator of composition for STA. We first make the assumption that the STA composed run completely independently from each other, i.e. they do not communicate between them. We prove that our definition satisfies indeed this independence assumption. Such an operator of composition is not very interesting as in general, systems do communicate. But it is a necessary first step. We then introduce the new model of interactive STA (ISTA) that will allow for interactions between the systems. We define an operator of composition in ISTA that will make synchronisations possible between the systems and that is built on the previous composition in STA. We end this thesis with the identification of a subclass of ISTA in which all the qualitative and quantitative results provided in this thesis can be applied, and which thus comes with the nice compositional framework defined in the model.
Document type :
Theses
Complete list of metadatas

Cited literature [62 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01696130
Contributor : Abes Star <>
Submitted on : Tuesday, January 30, 2018 - 10:25:28 AM
Last modification on : Thursday, September 12, 2019 - 3:31:31 AM
Long-term archiving on: Friday, May 25, 2018 - 10:19:19 AM

File

75316_CARLIER_2017_archivage.p...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01696130, version 1

Citation

Pierre Carlier. Verification of Stochastic Timed Automata. Modeling and Simulation. Université Paris-Saclay; Université de Mons, 2017. English. ⟨NNT : 2017SACLN058⟩. ⟨tel-01696130⟩

Share

Metrics

Record views

541

Files downloads

269