Scalable Parallel Model Checking via Monte-Carlo Tree Search

Abstract

The future of model checking lies in parallel and distributed computing, but parallel graph search algorithms tailored to directed model checking remains an underdeveloped area of research. In this work, we examine the application of parallel Monte Carlo Tree Search algorithms. We demonstrate how exploratory, randomly sampled rollouts of the search space, coordinated through a minimally communicating work-sharing protocol, can enable us to push the boundaries on the scope and scale of problems amenable to serial search.