Physics engine for robotics and AI

RaiSim is a physics engine designed for deep reinforcement learning and other robotics applications. Jemin Hwangbo started developing it at ETH Zurich and is now developing it at KAIST.


Our training environment is about two orders of magnitude faster than the competitions'. It can generate more than 100 billion samples on a single CPU in one day.

Reinforcement learning environment for robotics

  • YouTube

Robotics & Artificial Intelligence Lab

Department of Mechanical Engineering
Korea Advanced Institute of Science and Technology
Building N9, 4150
291 Deahak St.
Yuseong-gu, Deajeon, Korea

©2020 by KAIST RAI Lab. All rights reserved