"Scalable Reinforcement Learning for a Simulated Production Line"
Deep reinforcement learning has been shown to be able to solve tasks without prior knowledge of the dynamics of the problems. In this thesis the applicability of reinforcement learning on the problem of production planning is evaluated. Experiments are performed in order to reveal strengths and weaknesses of the theory currently available. Reinforcement learning shows great potential but currently only for a small class of problems. In order to use reinforcement learning to solve arbitrary or a larger class of problems further work needs be done. This thesis was written at Syntronic Software Innovations.
Keywords: Reinforcement learning, Machine learning, artificial neural networks, production planning