Abstract: This paper introduces a novel framework for addressing selective maintenance problems with imperfect repairs over a finite horizon. The proposed approach relies on two deep Q-learning ...