Building a world model verifier for robotic task verification
Coding agents, like Claude, are now able to write robot policies. They can look at a scene, solve IK, plan long-horizon tasks, and more. See Waddle Labs and…
Blog — build logs and research writeups by Nikhil Prabhu
Blogs
How I taught a humanoid to score a penalty kick
I was recently watching the World Cup and naturally wondered if we could get humanoids to play soccer. Locomotion policies have been improving, and the Unitree G1 was…
Can object detection make robot policies more robust?
I just worked on my first mini experiment, which I am dubbing ACT-YOLO. I want to know if visual input gets corrupted for policies like ACT, does giving…