Key points are not available for this paper at this time.
Abstract Humanoid robots are increasingly being developed for seamless interaction with humans in various domains. However, generating expressive and feasible motions for these robots remains a significant challenge due to their complex kinematic structures and physical constraints. We propose a robust and automated pipeline for motion retargeting that enables the generation of natural motions for diverse humanoid robots using various motion data sources. Our approach unifies different kinematic configurations into a single predefined rig and refines the motion trajectory, considering factors such as balance and contact. The retargeted motion is then fine-tuned to closely follow the source motion while adhering to the robot's physical limits. We demonstrate the effectiveness of our methodology through successful applications on 12 simulated robots and validation on three real robots. This work represents a significant step towards automating expressive motion generation for humanoid robots, enabling their deployment in various real-world scenarios.
Jeong et al. (Wed,) studied this question.