Paper on Deep RL for Valkyrie accepted to Humanoids 2018
Chuanyu Yang, Kai Yuan, Wolfgang Merkt, Taku Komura, Sethu Vijayakumar, Zhibin Li. “Deep Reinforcement Learning of Locomotion Skills for the Humanoid Valkyrie”, Proc. IEEE Intl. Conf. on Humanoid Robots (Humanoids 2018), Beijing (2018).
This paper presents a hierarchical control framework for Deep Reinforcement Learning that can learn a wide range of push recovery and balancing behavior, i.e., ankle, hip, foot tilting, and stepping strategies. The policy is trained in a realistic physics simulator using the robot model in a setup designed to be able to easily transfer and deploy synthesized control policies to real-world platforms. The advantage over traditional methods that integrate high-level planner and feedback control is that one single coherent policy network is generic for generating versatile, unprogrammed balancing and recovery motions against unknown perturbations at arbitrary locations (e.g., legs, torso). Furthermore, the proposed framework allows the policy to be learned with any state-of-the-art learning algorithm. By comparing the proposed approach with other methods in the literature, we found the performance of learning being similar in terms of disturbance rejection ability with additional benefits of generating generic and versatile behavior.
to be added.