Learning to Walk in Minutes Using Massively Parallel Deep RL

We present a training set-up that achieves fast policy generation for real-world robotic tasks by using massive parallelism on a single workstation GPU. The parallel approach allows training policies for flat terrain in under 4 minutes, and in 20 minutes for uneven terrain.

Paper accepted to CoRL 2021.
The corresponding code will be released soon.
Paper: https://arxiv.org/abs/2109.11978