Home▸Robotics & Kinematics▸Sim-to-Real Transfer: Domain Randomization

🎲 Sim-to-Real Transfer: Domain Randomization

Domain randomization vs. a single fixed-friction simulation: train a push policy either way, then test it on a real robot with unseen friction and see which one survives the reality gap.

Robotics & Kinematics3DAdvanced60 FPS

⚙ Under the hood

A pusher robot learns to push a box to a target either in one fixed-friction simulation or across thousands of randomized-friction episodes, then both policies are tested on a real robot with an unseen friction value, showing which one survives the reality gap.

Three.jsroboticssim-to-realdomain randomizationreality gapreinforcement learning

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install