ARISE-Initiative/robosuite

Image based RL for Robosuite

Closed this issue · 3 comments

Hi !
I was wondering if there are any repos that do Image-based online RL for robosuite. I saw the robosuite-benchmark repo and it seems to be using the low level dimensions for objects directly.

@snasiriany @amandlek do we have image based RL somewhere for robosuite?

We have image based online PPO RL (Diffusion/Gaussian/GMM) set up for Can/Lift/Square/Transport with DPPO. I think IBRL has Can/Square set up with Q-learning.

Amazing; thanks for sharing these, Allen! I'll pin this issue. Congrats on the awesome work :)