Learning to Walk with Hybrid Serial-Parallel Linkages: a Case Study on the Kangaroo Robot

Abstract

<div> Humanoid robots increasingly adopt hybrid serialparallel kinematics to improve structural stiffness, mass distribution, and impact robustness. However, these mechanisms introduce complexity associated with simulation and control, which impacts algorithms for Reinforcement Learning (RL)based locomotion. This paper presents a case study of an RL end-to-end pipeline that trains walking policies for Kangaroo, a 72 degrees of freedom biped whose legs contain several hybrid serial-parallel chains, without kinematic simplifications. Training is performed using the Isaac Lab framework, leveraging the Isaac Sim ™ built-in constraint capabilities. An ablation study on the observation state is carried out to find evidence in the use of redundant information from the measured state of the robot, i.e., using the passive and/or active joint measurements available in Kangaroo. A set of trained policies is validated in MuJoCo, demonstrating a degree of robustness to the Simto-Sim gap, provided that the equality-constraint stiffness and other simulation parameters are properly tuned. The closed-loop behaviors of the tested policies successfully transfer in most cases, despite differences in how contacts and constraints are modeled across the two simulators. Furthermore, we analyzed how minor differences in the action rate penalty weight used during training can deeply affect the locomotion stability of the resulting policies when deployed in a different simulation environment. </div>

Affiliated Institutions

• Université de Lorraine FR
• University College London GB
• Laboratoire Lorrain de Recherche en Informatique et ses Applications FR

Related Publications