experiment details

[ethan5722]

1. Whether the current state of the autonomous vehicle is used as model input ?
2. Have you done experiments with unsupervised latents loss ?

[liyingyanUCAS]

Sorry for the late reply. Here is the responses:

1. We don't use the ego status.
2. What do you mean about unsupervised latents loss?

[ethan5722]

I would like to ask whether your trajectory_prediction task and world_model task are trained together, or are they trained in two stages? Do you freeze the image encoder while train world_model?