Original title: Scheduling Distributed Flexible Assembly Lines using Safe Reinforcement Learning with Soft Shielding
Authors: Lele Li, Liyong Lin
In manufacturing, automated assembly lines boost productivity, but scheduling remains a challenge, especially for customized orders. This article introduces a solution—a reinforcement learning method—to manage flexible assembly line schedules in real-time. The approach streamlines the environment representation and action space, ensuring efficiency. Additionally, a soft shielding technique, employing Monte-Carlo tree search, tackles safety concerns linked to lengthy scheduling sequences. The method’s performance undergoes validation, affirming its potential in improving efficiency, reducing delays, and ensuring safety in assembly line scheduling. This innovation addresses critical challenges, paving the way for smoother, safer, and more efficient assembly line operations in manufacturing settings.
Original article: https://arxiv.org/abs/2311.12572