Reproduce “Half-Physics: Enabling Kinematic 3D Human Model with Physical Interactions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Li Siyao, Yao Feng, Omid Taheri, Chen Change Loy, Michael J. Black
The "half-physics" framework seamlessly integrates kinematic 3D human models, specifically SMPL-X, into physics simulators. This enables physically plausible interactions with environments and objects by transforming kinematic poses into directly applied velocities, thereby preserving original motion fidelity and operating in real-time without requiring training.
Key insights
- Directly translating discrete kinematic pose differences into velocities for a physics engine enables high-fidelity motion tracking and real-time, physically plausible interactions without extensive training.
- Disabling typical physics engine features like joint damping and internal forces, alongside implementing a passive joint stiffness compensation, is essential for maintaining kinematic fidelity while integrating with dynamic physical responses.
- A learning-free, velocity-driven method can effectively bridge the gap between kinematic control and physical simulation, offering an efficient and generalized alternative to computationally intensive RL-based physical trackers.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.