Checking Multi-Agent Systems Against Temporal-Epistemic Specifications

Abstract

This work is on methodologies for checking multi-agent systems against temporal-epistemic specifications. Since behaviors that comply with temporal-epistemic specifications in general (or theoretically) involve infinite sequences of actions of the involved agents, we must avoid checking such specifications based on such infinite sequences. This work at the theoretical side provides a bounded semantics for the temporal-epistemic specification formalism CTLK such that checking an infinite number of steps can be reduced to checking a finite number of steps, and at the practical side develops an approach based on the semantics and QBF-solving techniques for the verification purpose. The approach has been implemented and experimental data show that there exist verification problems that can be verified by this approach more efficiently than BDD based symbolic model checking.