Abstract

Pedestrian trajectory prediction is one of the important tasks required for autonomous navigation and social robots in human environments. Previous studies focused on estimating social forces among individual pedestrians. However, they did not consider the social forces of groups on pedestrians, which results in over-collision avoidance problems. To address this problem, we present a Disentangled Multi-Relational Graph Convolutional Network (DMRGCN) for socially entangled pedestrian trajectory prediction. We first introduce a novel disentangled multi-scale aggregation to better represent social interactions, among pedestrians on a weighted graph. For the aggregation, we construct the multi-relational weighted graphs based on distances and relative displacements among pedestrians. In the prediction step, we propose a global temporal aggregation to alleviate accumulated errors for pedestrians changing their directions. Finally, we apply DropEdge into our DMRGCN to avoid the over-fitting issue on relatively small pedestrian trajectory datasets. Through the effective incorporation of the three parts within an end-to-end framework, DMRGCN achieves state-of-the-art performances on a variety of challenging trajectory prediction benchmarks.

Talk

Motivation

Over-Smoothing and Biased Weighting Problems

Even though graph-based approaches can represent arbitrary structures well, two problems limit their applicability for pedestrian trajectory prediction. First of all, they suffer from over-smoothing problems on node features. When constructing pedestrian graphs for crowded environments, the features are smoothed by the aggregation of lots of nodes. Another problem comes from high-order social relations. The weight bias problem occurs when powering the adjacency matrix for multi-scale aggregation operations. The edge weight increases exponentially when there are lots of strong connections on the k-hop neighborhoods. In this work, we present a novel disentangled multi-scale aggregation of social relations on a weighted graph.

Prediction Error Accumulation

Most pedestrian trajectory prediction models have a common problem, in that prediction errors are accumulated as the sequences become longer. Particularly, this problem arises when a person turns around an obstacle or another person. In this work, we tackle this problem by proposing a novel global temporal aggregation (GTA) which learns to compensate for the accumulated error. GTA takes each pedestrian’s trajectory as input, and outputs a single feature vector which is added to the initial prediction.

Qualitative Results

BibTeX

@article{bae2021dmrgcn,
  title={Disentangled Multi-Relational Graph Convolutional Network for Pedestrian Trajectory Prediction},
  author={Bae, Inhwan and Jeon, Hae-Gon},
  journal={Proceedings of the AAAI Conference on Artificial Intelligence},
  year={2021}
}