some questions of Observation

[yangshuocun]

Could you please explain the details of the Observation tensor as it is defined in the file het_control_mlp_empirical.py?I hope this message finds you well. I am writing to inquire about the structure of the observation tensor in your work, which is shaped as [*batch, n_agents, n_features]. Could you please elaborate on the specific components that constitute the n_features dimension?
Thank you for your time and for sharing your insightful research.

[matteobettini]

Hello,

the n_features will correspond to the values returned by the observation function of the corresponding VMAS scenario.

Take dispersion for example, you can see the values here https://github.com/proroklab/VectorizedMultiAgentSimulator/blob/acd9b7aca3ca5718f58f03a993c5ef3e920c6af9/vmas/scenarios/dispersion.py#L136 in the scenario file