Physics-Based Motion Tracking of Contact-Rich Interacting Characters

2026-04-09

AI Summary

This work addresses the instability of existing physics-driven motion tracking methods when handling contact-rich multi-character interactions, which stems from unstable single-character control and high control demands induced by contact force transmission. To overcome these limitations, the study introduces Progressive Neural Networks (PNN) to this task for the first time, proposing a robust multi-character motion tracking approach. The method employs a mixture-of-experts architecture to automatically allocate training samples and jointly learn interaction skills of varying complexity, integrated with a contact-aware, physics-driven control strategy. Notably, it eliminates the need for manual curriculum scheduling and significantly enhances both tracking stability and training efficiency in dense interaction scenarios. Comprehensive qualitative and quantitative evaluations demonstrate clear superiority over current state-of-the-art methods.

Abstract

Motion tracking has been an important technique for imitating human-like movement from large-scale datasets in physics-based motion synthesis. However, existing approaches focus on tracking either single character or a particular type of interaction, limiting their ability to handle contact-rich interactions. Extending single-character tracking approaches suffers from the instability due to the challenge of forces transferred through contacts. Contact-rich interactions requires levels of control, which places much greater demands on model capacity. To this end, we propose a robust tracking method based on progressive neural network (PNN) where multiple experts are specialized in learning skills of various difficulties. Our method learns to assign training samples to experts automatically without requiring manually scheduling. Both qualitative and quantitative results show that our method delivers more stable motion tracking in densely interactive movements while enabling more efficient model training.

Problem

• motion tracking
• contact-rich interaction
• physics-based animation
• multi-character interaction
• force transfer

Innovation

• physics-based motion tracking
• contact-rich interaction
• progressive neural network
• multi-expert learning
• automatic sample assignment