Hugging Face Releases CoDance Framework: Revolutionizing Human-Humanoid Interaction Learning

Summary:Hugging Face has introduced CoDance, a novel framework for learning reactive and compliant human-humanoid interactions from video. By converting videos of two human dancers into robot reference trajectories and a moving partner, CoDance introduces multi-link compliance augmentation to transform kinematic demonstrations into force-aware training data. Simulation experiments demonstrate that CoDance enables robots to adapt to partner movements while maintaining a stable locomotion style, and physi

---

Key Breakthroughs

Hugging Face's research team has introduced CoDance, a novel framework designed to tackle the complexities of human-humanoid interaction. The core innovations of CoDance include:

• Video-to-robot motion conversion: By analyzing videos of two human dancers, CoDance converts human movements into robot reference trajectories and the motions of a moving partner.
• Multi-link compliance augmentation: This technique transforms kinematic demonstrations into force-aware training data, enabling the robot to perceive and respond to interaction forces.
• Maintaining locomotion style and compliance: In simulations, CoDance-trained policies adapt to changes in the partner's movements while preserving the original locomotion style and responding compliantly to physical interactions.

Technical Highlights

1. Single video input: CoDance requires only a single video of two human dancers to generate interaction trajectories for the robot and the partner, reducing the need for extensive training data.
2. Compliance augmentation: By adjusting the robot reference trajectory under structured forces at both hands, CoDance achieves perception and response to physical forces.
3. Simulation and physical testing: In simulations, CoDance reproduces approximately 80% of the wrist displacement encoded by the augmented demonstrations. On a physical humanoid, CoDance enables sustained two-handed dancing with a human partner, including repeated transitions between forward and backward motions.

Industry Impact

The release of CoDance marks a significant milestone in the field of human-humanoid interaction. Its potential applications are vast, including:

• Human-robot collaboration: In manufacturing and service industries, robots can collaborate more naturally with humans.
• Rehabilitation and assistance: In rehabilitation therapy, robots can better adapt to patients' movement patterns.
• Entertainment and education: In dance instruction and interactive entertainment, robots can provide more realistic interaction experiences.

Developer Recommendations

• Experimentation and extension: Developers can experiment with applying CoDance to other types of interaction scenarios, such as handshakes, hugs, etc.
• Multimodal fusion: Combining CoDance with data from other sensors (e.g., force sensors, IMUs) can further enhance the accuracy and safety of interactions.
• Real-time optimization: For real-time applications, optimizing CoDance's inference speed can enable smoother interaction experiences.

Conclusion

CoDance demonstrates the significant potential of AI in the field of robot interaction, opening new directions for future human-robot collaboration.

---

Source: Hugging Face Daily Papers (2026-10-04)

— END —

Tags: #Hugging Face Follow #Intelligent Agent Follow #Human-Robot Interaction Follow #Robotics Follow #AI Framework Follow