This figure shows the difference between prior methods and our proposed scheme. (IMAGE)
Caption
Comparison between prior methods and our proposed scheme. Existing approaches, illustrated in (a), predict trajectories under the idealized assumption of zero latency. In practice, however, the forecasting module in autonomous driving systems inevitably experiences delays, rendering predictions within this period obsolete since the movement of the target agent has already taken place. Our approach, depicted in (b), introduces a novel prediction paradigm that explicitly accounts for latency and leverages this seemingly useless period as auxiliary information to enhance predictive performance.
Credit
Communications in Transportation Research
Usage Restrictions
Credit must be given to the creator.
License
CC BY