Paper Image

Preventing unsafe vehicle trajectory predictions

Published on:

18 October 2023

Primary Category:

Robotics

Paper Authors:

Abhishek Vivekanandan,

Ahmed Abouelazm,

Philip Schörner,

J. Marius Zöllner

Bullets

Key Details

A refinement layer generates physically feasible trajectories adhering to map constraints

Attention mechanisms learn interactions between trajectories and goal lanes

The method achieves high accuracy while ensuring safety via on-road predictions

Extensive experiments validate the approach outperforms prior methods

AI generated summary

Preventing unsafe vehicle trajectory predictions

This paper presents a two-stage model for predicting future vehicle trajectories that ensures safety and feasibility. A refinement layer first generates physically possible trajectories using map data and vehicle kinematics. Then a neural network learns trajectory probabilities conditioned on the refined set, incorporating cross-attention between trajectories and goal lanes. This method achieves state-of-the-art performance while eliminating almost all unsafe, off-road predictions.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up