AFRL-AFOSR-JP-TR-2017-0017 Independence-based Optimization of Epistemic Model Checking

Epistemic model checking is an automated verification methodology that enables multi-agent and concurrent systems and models to be verified with respect to properties concerning the information possessed by the agents, and how this information changes over time. The project investigated an approach the optimization of epistemic model checking using reasoning about independencies detected by means of a static analysis technique. A theoretical basis for the optimization was developed, extending prior work on conditional independencies from the Bayesian Net literature. The resulting algorithm was implemented in the epistemic model checker MCK. Experiments on a number of benchmarks for epistemic model checking confirm that the optimization results in significant improvements in computation time, of as much as four orders of magnitude in some cases. It also enables model checking problems of significantly larger scale to be handled.