Reinforcement learning methods based on GPU accelerated industrial control hardware

Abstract Reinforcement learning is a promising approach for manufacturing processes. Process knowledge can be gained automatically, and autonomous tuning of control possible. However, the use reinforcement in production environment imposes specific requirements that must met successful application. This article defines those evaluates three methods to explore their applicability. The results show convolutional neural networks are computationally heavy violate real-time execution requirements. A new architecture presented validated allows using GPU-based hardware acceleration while meeting