Why Historical Predicted Trajectory?

[ghai2023]

As is mentioned in the paper, historical predicted trajectory is leveraged as condition input to maintain temporal consistency. This seems to be a pretty good idea. Could you please help to answer some of the questions?

1. why did you use historical predicted trajectory instead of historical expert trajectory?
2. Is the ego or agents' historical predicted trajectory utilized as condition input? If it was also applied to agents, then was it single modal or multi-modal predicted trajectory of one agent?
3. How did you get the historical predicted trajectory? I mean ,what was the approach?