Does “massively parallel simulation” help advance Reinforcement Learning?

Re:Does “massively parallel simulation” help advance Reinfor... by hcrisp on Friday December 09, 2022 @07:09PM (#33928218)

It's not new. A paper in 2021 [0] showed you can train a quadruped robot to walk in minutes using parallel simulation in GPU, and then deploy it on the physical robot. Being parallel it is faster, and more so if on GPU. But sim-to-reality transfer is still a concern, and the architecture doesn't help if there are sparse rewards.

[0] https://arxiv.org/abs/2109.11978?context=cs.LG

[ Parent | on HN ]

Re:Does “massively parallel simulation” help advance Reinfor... by yanglet on Saturday December 10, 2022 @03:36PM (#33936548)

Thanks for the information and sharing the paper. I also read it. Agree with you that the "simulation-to-reality gap" would be more critical to the reinforcement learning community.

[ Parent | on HN ]