Machine Learning - Deep Reinforcement Learning

In order to build fast prototypes we are using Python in combination with NumPy tens of flow PyTorch and so on whatever matches best but I think interest the interesting part is when we want to get real like to test our models on real hardware on hydraulics then we have to translate our algorithms to ccode because otherwise wouldn't run on established PLCs at least the PLCs that we use in hydraulics. The data hungariness comes relief from the many different working conditions and working points that we are talking about.