Symbolic Model Checking of Stochastic Reward Nets

This paper describes a symbolic model checking approach for the Continuous Stochastic Reward Logic (CSRL) and stochastic reward nets, stochastic Petri nets augmented with rate rewards. CSRL model checking requires the computation of the joint distribution of time and accumulated reward, which is done by Markovian approximation. An implementation is available in the model checker MARCIE. It applies a symbolic on-the-fly computation of the underlying matrix using Interval Decision Diagrams. The result is a multi-threaded model checker which is able to evaluate CSRL formulas for models with millions of states depending on the required accuracy.