[Question] Observation space formulation

[VineetTambe]

Hey Authors,

Thanks for releasing the code!
I was going through the code and found option to use Maximal coordinates (explicit 6 DOF pose of each link) and Minimal coordinates (joint angles of humanoid) to construct the observation [humanoid_odestrian_terrain.py].

Do you have any insights into which representation gives better training results or helps in improving downstream tasks?

Thanks!