Sim-to-Real of Humanoid Locomotion Policies via Joint Torque Space Perturbation Injection
This addresses the challenge of training robust control policies for humanoid robots in real-world environments, representing an incremental improvement over existing domain randomization methods.
The paper tackles the problem of sim-to-real transfer for humanoid locomotion by proposing a method that injects state-dependent perturbations into joint torque space during training, which results in policies achieving greater robustness against unseen reality gaps.
This paper proposes a novel alternative to existing sim-to-real methods for training control policies with simulated experiences. Prior sim-to-real methods for legged robots mostly rely on the domain randomization approach, where a fixed finite set of simulation parameters is randomized during training. Instead, our method adds state-dependent perturbations to the input joint torque used for forward simulation during the training phase. These state-dependent perturbations are designed to simulate a broader range of reality gaps than those captured by randomizing a fixed set of simulation parameters. Experimental results show that our method enables humanoid locomotion policies that achieve greater robustness against complex reality gaps unseen in the training domain.