Abstract
Pedestrian trajectory prediction is an essential task in robotic applications such as autonomous driving and robot navigation. State-of-the-art trajectory predictors use a conditional variational autoencoder (CVAE) with recurrent neural networks (RNNs) to encode observed trajectories and decode multi-modal future trajectories. This process can suffer from accumulated errors over long prediction horizons (>=2 seconds). This paper presents BiTraP, a goal-conditioned bi-directional multi-modal trajectory prediction method based on the CVAE. BiTraP estimates the goal (end-point) of trajectories and introduces a novel bi-directional decoder to improve longer-term trajectory prediction accuracy. Extensive experiments show that BiTraP generalizes to both first-person view (FPV) and bird’seye view (BEV) scenarios and outperforms state-of-the-art results by ~10-50%. We also show that different choices of nonparametric versus parametric target models in the CVAE directly influence the predicted multi-modal trajectory distributions. These results provide guidance on trajectory predictor design for robotic applications such as collision avoidance and navigation systems.
Links
IEEE Xplore
arXiv
PDF
GitHub: BiTraP code
Citation
Plain Text:
Y. Yao, E. Atkins, M. Johnson-Roberson, R. Vasudevan and X. Du, “BiTraP: Bi-directional Pedestrian Trajectory Prediction with Multi-modal Goal Estimation,” in IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 1463-1470, April 2021.
BibTex:
@ARTICLE{yao2021bitrap,
author={Y. {Yao} and E. {Atkins} and M. {Johnson-Roberson} and R. {Vasudevan} and X. {Du}},
journal={IEEE Robotics and Automation Letters},
title={BiTraP: Bi-directional Pedestrian Trajectory Prediction with Multi-modal Goal Estimation},
volume={6},
number={2},
pages={1463-1470},
year={2021}}