TorchSharp PPO re-wired to use Unity's sensors. #1371
asieradzk
started this conversation in
Show and tell
Replies: 1 comment
-
|
My graphs look less impressive and they always get deformed around 20k episodes.... |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
As I am working on Unity plugin for my DRL library I thought to have a go re-wiring sensors that were made for ml-agents. Actually quite simple and works like a charm, I simply wrap around them and expose observations in friendly manner.
Video for your leisure of PPO having a field day with PushBlock environment. It is intentionally made more difficult by only providing observations via sensor. So agent doesn't know where it is (rays are too short and not omni-directional) or where block is - partially observable environment.
rlmatrix2.mp4
Beta Was this translation helpful? Give feedback.
All reactions