Reinforcement Learning and Intralogistics: Soft Actor Critic for Maples Navigation in Warehouses

Experience firsthand how we apply state-of-the-art Soft Actor Critic reinforcement learning algorithms in an intralogistics setting using PyTorch, OpenAI Gym and the Nvidia Isaac SDK. Our goal is to share with you how we manage to navigate a robot to a target area, based solely on Lidar or camera data, completely without the use of a map. We also talk about the architecture of our neural networks and show you some tricks on how to combine Nvidia Isaac and OpenAI. Furthermore we will show how we have parallelized and accelerated the training phase of our algorithms by utilizing an NVIDIA DGX.