Not Only Rewards But Also Constraints: Applications on Legged Robot Locomotion

arXiv:2308.12517v3 Announce Type: replace-cross
Abstract: Several earlier studies have shown impressive control performance in complex robotic systems by designing the controller using a neural network and training it with model-free reinforcement learning. However, these outstanding controllers with natural motion style and high task performance are developed through extensive reward engineering, which is a highly laborious and time-consuming process of designing numerous reward terms and determining suitable reward coefficients. In this work, we propose a novel reinforcement learning framework for training …

applications arxiv constraints cs.ai cs.lg cs.ro legged robot robot type