Poster 30: Probabilistic Occupancy Prediction for Autonomous Vehicles
(Room Palms Ballroom)
16 Jul 19
5:30 PM
-
7:00 PM
Tracks:
Vehicle Automation Technology
The poster presents a new approach to predict the occupancy areas of surrounding vehicles for autonomous driving. These occupancy areas should be as small as possible to provide enough drivable space for the ego vehicle but simultaneously large enough to almost certainly include the real movement. For this purpose, a probabilistic overapproximativ prediction of the driver behavior is calculated, which results in the probability of presences. The predicted occupancy area can be determined on the basis of this probability as a function of an defined accepted risk. In the first step a physical model is used to determine the possible presence locations. The subsequent assessment of the movement options based on the statistically representative driver behavior, static objects and dynamic objects allows determining the probabilities of presences. Finally, a real world data assessment confirms that this new approach reduces the occupancy area size while keeping the error rate low.