Ways of integrating already existing knowledges into SAC training?

[famora2]

Hi,
I manually controlled the ant robot from this example (https://github.com/NVIDIA-Omniverse/IsaacGymEnvs/blob/main/isaacgymenvs/tasks/ant.py) and recorded the corresponding joint angle values.
I would like to know whether there is a way of integrating this per-existing knowledge into the agent training (such as the imitation learning from SB3: (https://imitation.readthedocs.io/en/latest/algorithms/gail.html)).

[Denys88]

We didn't try to add any IL papers yet because the main goal was the training from the scratch. But we can consider GAIL.