A Deep Reinforcement Learning Approach for the Hoist Scheduling Problem in Electroplating Lines: A Proof of Concept

The Hoist Scheduling Problem (HSP) arises in electroplating lines, where material handling equipment must transport workpieces through sequences of chemical baths within strict time windows. The problem is NP-hard, and its computational complexity renders traditional optimization methods impractical for industrial applications, as solution times typically exceed operational constraints. This paper presents a Deep Reinforcement Learning (DRL) approach to the HSP, employing Deep Q-Networks (DQN) as a proof of concept. The electroplating line is modelled as a Markov Decision Process in which states encode the system configuration, actions represent feasible hoist movements, and rewards quantify scheduling efficiency. Experimental evaluation on a simplified HSP instance demonstrates stable learning dynamics and competitive performance. This work does not aim at outperforming state-of-the-art optimization methods, but rather at assessing the feasibility of framing the Hoist Scheduling Problem as a Markov Decision Process and learning a control policy via Deep Reinforcement Learning. Experimental results on a simplified instance show stable learning behavior and suggest the potential of DRL-based approaches to handle variations in initial conditions and processing times, paving the way for adaptive scheduling strategies in electroplating systems.