Have you tried this method to train the standard 3d humanoid?

huangwl18 / modular-rl

[ICML 2020] PyTorch Code for "One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Control"

https://huangwl18.github.io/modular-rl/

Other

217 stars 34 forks source link

Have you tried this method to train the standard 3d humanoid? #5

Closed R-Ceph closed 4 years ago

R-Ceph commented 4 years ago

How many steps does it take to train a 3D model in this way? Your idea is very innovative, and I want to use this to improve my policy.

huangwl18 commented 4 years ago

Thanks for the interest in the paper!

We didn't try to train the policy on a 3d humanoid as most of the Gym MuJoCo locomotion agents are in 2d and the goal is to investigate the possibility of training a single policy for multiple agent morphologies. However, it is expected to work well in 3d too, but one might need to modify the dimensionality of input/output correspondingly.