GentleHumanoid: Learning Upper-body Compliance for Contact-rich Human and Object Interaction

TL;DR

GentleHumanoid tackles the challenge of safe, compliant physical interaction for humanoids by embedding impedance control into a whole-body motion-tracking policy. It introduces a unified spring-based interaction model that handles both resistive and guiding contacts, with adaptive force thresholds to enforce safety, and trains the policy via a PPO-based teacher–student framework using diverse human-motion datasets. Across simulation and real-world tests on the Unitree G1, it achieves lower peak contact forces and smoother, more natural interactions in tasks like hugging, sit-to-stand, and fragile-object handling, outperforming baselines. The work advances practical human–robot collaboration by enabling coordinated upper-body compliance, vision-assisted personalization, and sim-to-real transfer, paving the way for safer assistive and interactive humanoids in real environments.

Abstract

Humanoid robots are expected to operate in human-centered environments where safe and natural physical interaction is essential. However, most recent reinforcement learning (RL) policies emphasize rigid tracking and suppress external forces. Existing impedance-augmented approaches are typically restricted to base or end-effector control and focus on resisting extreme forces rather than enabling compliance. We introduce GentleHumanoid, a framework that integrates impedance control into a whole-body motion tracking policy to achieve upper-body compliance. At its core is a unified spring-based formulation that models both resistive contacts (restoring forces when pressing against surfaces) and guiding contacts (pushes or pulls sampled from human motion data). This formulation ensures kinematically consistent forces across the shoulder, elbow, and wrist, while exposing the policy to diverse interaction scenarios. Safety is further supported through task-adjustable force thresholds. We evaluate our approach in both simulation and on the Unitree G1 humanoid across tasks requiring different levels of compliance, including gentle hugging, sit-to-stand assistance, and safe object manipulation. Compared to baselines, our policy consistently reduces peak contact forces while maintaining task success, resulting in smoother and more natural interactions. These results highlight a step toward humanoid robots that can safely and effectively collaborate with humans and handle objects in real-world environments.

Figures (6)

• Figure 1: GentleHumanoid learns a universal whole-body control policy with upper-body compliance and tunable force limits. It enables: (a) sit-to-stand assistance, where the robot provides support across multiple links (hand, elbow, and shoulder); (b) handshaking with a 5 N force limit, allowing the robot’s hand to move naturally with the human’s; (c) autonomous shape-aware hugging, where the robot adapts its posture to the partner’s body shape (estimated from camera input) for a comfortable embrace; and (d) balloon handling, showing safe object manipulation where baselines fail.
• Figure 2: Overview framework. (a) Reference dynamics: impedance-based dynamics integrate driving forces (for motion tracking) and interaction forces (for compliant contact), producing reference link (on the shoulders, elbows and hands) positions and velocities. (b) Training: the policy receives proprioception, privileged observations, and target motions, and is optimized using rewards that compare simulated states $(\bm{x}^{\text{sim}}, \dot{\bm{x}}^{\text{sim}})$ to reference dynamics $(\bm{x}^{\text{ref}}, \dot{\bm{x}}^{\text{ref}})$. (c) Deployment: the trained GentleHumanoid policy is applied to real-world tasks, including vision-based autonomous hugging and other human–robot interaction scenarios, enabling safe and compliant behaviors such as hugging, sit-to-stand assistance, and handling large deformable objects.
• Figure 3: Interaction force distributions across upper-body links. Probability densities of force magnitudes are shown for the right shoulder (left), right elbow (middle), and right hand (right). Insets (top right) illustrate the corresponding force directions on a sphere.
• Figure 4: Forces applied by different upper-body links under external interaction. Force profiles over time are shown for the right hand (left), right elbow (middle), and right shoulder (right). Compared to baselines (Vanilla-RL and Extreme-RL), GentleHumanoid maintains lower and more stable force levels across all links, showing safer and more compliant responses during contact.
• Figure 5: Comparison of interaction forces across policies. Top: GentleHumanoid with tunable force limits, which maintains safe interaction by keeping contact forces within specified thresholds across different postures. Bottom: baseline methods, Vanilla-RL and Extreme-RL, exhibit less consistent compliance, with higher peak forces or oscillatory responses. Force gauge readings (N) are highlighted for clarity.