PHYSICS SIM LAB · TECH BLOG
Building worlds where robots learn.
We build training infrastructure for Physical AI with physics-based simulation and synthetic data.
Accurate physics, robot control, and verifiable data. We share what we learn through implementation and experiments.
Company websitePosts
- ENGINEERING NOTE
Wuji Hand Tactile Data: From Hardware Interfaces to MuJoCo 4.0 Forces
Mapping Wuji Hand's tactile contract into MuJoCo 4.0.0: contact forces, 176 sensing points, 15 resultant components, force images, and their validation limits.
- ENGINEERING NOTE
Paper Folding and Pressing in MuJoCo 4.0: Forming a Curve, Remembering a Crease
A MuJoCo CPU experiment that records only deep pressing and releases the grip according to supported contact travel. Videos and measurements show IPC retries, crease retention after release, and remaining thickness errors.
- ENGINEERING NOTE
Grasping a Kettle Handle with Wuji Hand 2.1 · Design and Validation
A simulation experiment measuring contact forces, tilt, and return error as four fingers of Wuji Hand 2.1 wrap around a 500 g kettle's handle and lift it.
- ENGINEERING NOTE
Putting Clothes on a Hanger in Newton · VBD and Style3D
Comparing shirt folds on the same mesh in Newton VBD and Style3D, then validating hanger contact and grasp release with official women's T-shirt and sweatshirt assets.
- ENGINEERING NOTE
Newton IPC: Independent Implementation and Porting Guide
Implementation and porting of a Newton-connected IPC solver that jointly resolves triangular film and spherical contents, with inputs, energies, collision handling, and validation from a 481-vertex bag experiment.
