Improving the Model Checking of Strategies under Partial Observability and Fairness Constraints

Abstract

Reasoning about strategies has been a concern for several years, and many extensions of Alternating-time Temporal Logic have been proposed. One extension, ATLK irF , allows the user to reason about the strategies of the agents of a system under partial observability and unconditional fairness constraints. However, the existing model-checking algorithm for ATLK irF is inefficient when the user is only interested in the satisfaction of a formula in a small subset of states, such as the set of initial states of the system. We propose to generate fewer strategies by only focusing on partial strategies reachable from this subset of states, reducing the time needed to perform the verification. We also describe several practical improvements to further reduce the verification time and present experiments showing the practical impact of the approach.