Is there anyway to know the each dimension meaning of observation and action？

[zoeyuchao]

Is there a way or documentation to explain the physical meaning of each dimension in obs and action?

[larsrpe]

All of the data in the rl-pipeline is lifted by a "batch-dim" i.e the number of parallel environments.

So say for each environment you define 12 actions, e.g joint-positions on a quadruped, the action tensor would be of size (num_envs,12). Same goes for literally everything else in the pipeline!

Hope this helps!