SoftMimic: Learning Compliant Whole-body Control from Examples

2025-10-20

AI Summary

Humanoid robots often exhibit rigid trajectory tracking and poor compliance when learning from demonstration, limiting their ability to adaptively respond to external disturbances during physical interaction. Method: This paper proposes a compliant whole-body control framework that integrates inverse kinematics (IK)-enhanced data generation with reinforcement learning (RL). IK is employed to synthesize motion datasets explicitly encoding compliant responses, guiding policy learning toward contact-adaptive joint impedance modulation and coordinated whole-body dynamics—beyond mere trajectory tracking. An end-to-end RL controller is then trained for multi-task generalization. Contribution/Results: To our knowledge, this is the first approach to explicitly embed compliance priors into the imitation learning pipeline, enabling unified balance maintenance, disturbance rejection, and safe physical interaction from a single demonstrated motion. Evaluations in simulation and on a physical humanoid platform demonstrate significant improvements in stability, robustness, and safety within human–robot coexistence scenarios.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

We introduce SoftMimic, a framework for learning compliant whole-body control policies for humanoid robots from example motions. Imitating human motions with reinforcement learning allows humanoids to quickly learn new skills, but existing methods incentivize stiff control that aggressively corrects deviations from a reference motion, leading to brittle and unsafe behavior when the robot encounters unexpected contacts. In contrast, SoftMimic enables robots to respond compliantly to external forces while maintaining balance and posture. Our approach leverages an inverse kinematics solver to generate an augmented dataset of feasible compliant motions, which we use to train a reinforcement learning policy. By rewarding the policy for matching compliant responses rather than rigidly tracking the reference motion, SoftMimic learns to absorb disturbances and generalize to varied tasks from a single motion clip. We validate our method through simulations and real-world experiments, demonstrating safe and effective interaction with the environment.

Problem

• Learning compliant whole-body control from human motions
• Solving stiff and unsafe robot behavior during imitation
• Enabling compliant response to external forces while maintaining balance

Innovation

• Learns compliant control from augmented motion dataset
• Uses reinforcement learning with compliant response rewards
• Generates feasible motions via inverse kinematics solver