Algorithms for Deciding the Safety of States in Fully Observable Non-deterministic Problems: Technical Report

Researchers propose a new algorithm, iPI, for deciding the safety of states in fully observable non-deterministic problems, achieving a polynomial worst-case runtime while matching the best-case performance of TarjanSafe, a previously effective but exponentially slow algorithm. iPI outperforms a linear-time alternative in practice, demonstrating its efficiency and scalability.