Combining Learning-Based Locomotion Policy With Model-Based Manipulation for Legged Mobile Manipulators

Deep reinforcement learning produces robust locomotion policies for legged robots over challenging terrains. To date, few studies have leveraged model-based methods to combine these skills with the precise control of manipulators. Here, we incorporate external dynamics plans into learning-based mobile manipulation. We train base policy by applying a random wrench sequence on robot in simulation and add noisified prediction observations. The then learns counteract partially-known future disturbance. sequences are replaced generated from model predictive enable deployment. show zero-shot adaptation manipulators unseen during training. On hardware, demonstrate stable wrench.