Section: New Results

Stochastic extension of MARTE/CCSL for CPS modeling

Participant : Frédéric Mallet.

This work was conducted during the sabbatical period of Frédéric Mallet at ECNU Shanghai, in the context of the associated team FM4CPS ( 9.4.1.1 ).

As a declarative language, CCSL allows the specification of causal and temporal properties of systems expressed as constraints in a specific syntax. While each constraint reduces the set of possible behaviors, there may still be multiple (schedule) solutions, or none at all. When several solutions remain feasible, our TimeSquare tool allows to set up a resolving policy, to choose whether we want to attempt exploring exhaustively all these solutions, or else narrow the solution space according to an auxiliary criterion.

The extension of CCSL with stochastic features and probabilistic information is meant to help provide such an additional criterion, while modeling temporal constraints on the environment which are not necessarily well-known or controllable, specially in the domain of Cyber-Physical Systems. Then, such features should help reducing the set of possible behaviors, narrowing for instance to the most likely ones (in a formal quantitative meaning).

We are currently relying on UPPAAL SMC (Stochastic Model-Checking) toolset as prototype analyzer for the resulting specifications.