Selective Social-Interaction via Individual Importance for Fast Human Trajectory Prediction

TL;DR

This paper introduces an importance-based selection architecture for human trajectory prediction, utilizing a differentiable sampling method to improve speed while maintaining accuracy.

Contribution

It proposes the Importance Estimator module and Gumbel Softmax sampling to efficiently select relevant neighbors for trajectory prediction.

Findings

Speeds up trajectory prediction process

Maintains competitive accuracy on JRDB dataset

Effective neighbor selection improves prediction efficiency

Abstract

This paper presents an architecture for selecting important neighboring people to predict the primary person's trajectory. To achieve effective neighboring people selection, we propose a people selection module called the Importance Estimator which outputs the importance of each neighboring person for predicting the primary person's future trajectory. To prevent gradients from being blocked by non-differentiable operations when sampling surrounding people based on their importance, we employ the Gumbel Softmax for training. Experiments conducted on the JRDB dataset show that our method speeds up the process with competitive prediction accuracy.