Overall Objectives
Research Program
Application Domains
New Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
XML PDF e-pub
PDF e-Pub

Section: New Results

Timed, Probabilistic, and Stochastic Extensions

Model Checking for Extended PCTL

Participants : Hubert Garavel, Radu Mateescu, Jose Ignacio Requeno.

In the context of the SENSATION project (see § ), we study the specification and verification of quantitative properties of concurrent systems.

In 2014, we defined an extension of PCTL (Probabilistic Computation Tree Logic[49] with the manipulation of data values and actions. This logic is interpreted on extended DTMCs (Discrete-Time Markov Chains) containing visible transitions, labeled with channel names and data values, in addition to probabilistic transitions. Extended PCTL makes possible the specification of temporal properties involving discrete time, probabilities, and data values.

We devised a prototype model checker for extended PCTL in the form of an XTL library describing the denotational semantics of all PCTL operators (both primitive and derived ones), accompanied by external C code implementing the algorithms for LTS exploration and numerical computation of probabilities. The high-level programming language constructs of XTL (iterators, sets in comprehension, parameterized macro-definitions) allowed us to easily implement the advanced features (filters on arithmetic and logical operators, computation of probabilities, experiments over data series, etc.) of established probabilistic model checkers, such as PRISM  [54] . Also, the manipulation of data values in XTL allows one to specify properties in which probabilities and discrete time deadlines depend on the values of state variables, a feature currently not provided by PRISM.

To experiment and cross-check our extended PCTL library w.r.t. PRISM, we developed an automated translator from the (state-based) DTMCs used by PRISM into the (action-based) DTMCs in BCG format used by CADP. State information is represented by means of special self-looping transitions containing the values of state variables, which are properly handled during the evaluation of probabilistic temporal operators.

The experiments we performed with our extended PCTL library on various examples of DTMCs (produced from communication protocols, chemical reactions, hazard games, etc.) showed a performance comparable to (explicit-state) PRISM for pure PCTL formulas.

Furthermore, in addition to many bug fixes, the XTL compiler and its XTL_EXPAND preprocessor have been strengthened to better detect and report potential mistakes in source XTL specifications. In particular, vacuity checks have been introduced, which warn the user when no label in a BCG graph has the right number of fields or the appropriate field types to satisfy an XTL label match expression (previously, this expression would silently evaluate to false).

The type checking system of XTL and its list of predefined functions have been extended to support the new Natural and Raw types of the BCG format, and to properly distinguish between Natural and Integer values, and Raw and String values, while achieving a high degree of backward compatibility. In particular, XTL now uses type information from the BCG labels to better solve overloading in label offers, so that certain XTL programs that were formerly invalid are now accepted. Finally, it is now possible to use the predefined types and functions of XTL when defining temporal operators directly using external C code.