Chess pieces are heavy and our arms are weak! Luckily, we have strong robots and smart
Robot Foundation Models (RFMs). So we built a voice-controlled robot arm that moves your chess pieces for you!
In this showcase, a
UR5 robot executes chess moves using NVIDIA’s
GR00T N1.5. Instead of relying on real world data collection, we generate expert demonstrations entirely in simulation. The full setup - robot, board, and pieces - is recreated in
NVIDIA Isaac Sim, where high quality demonstrations are recorded and used to fine-tune the model.
The challenge is bringing these learned skills into the real world. Differences in perception, calibration, and physics make
sim2real transfer a non-trivial problem.
This system demonstrates how far a simulation first approach can go in enabling precise robotic manipulation.
Read more
→