Are socially-aware trajectory prediction models really socially-aware?

TL;DR

This paper introduces a novel attack method to evaluate and improve the social understanding of neural network-based trajectory prediction models, revealing their limitations in collision avoidance and robustness.

Contribution

We propose a socially-attended attack that exposes weaknesses in current trajectory prediction models and can be used to enhance their social understanding.

Findings

Current models have limited social understanding in collision avoidance.

Our attack effectively exposes vulnerabilities in recent trajectory predictors.

Using our attack can improve the social awareness of state-of-the-art models.

Abstract

Our field has recently witnessed an arms race of neural network-based trajectory predictors. While these predictors are at the core of many applications such as autonomous navigation or pedestrian flow simulations, their adversarial robustness has not been carefully studied. In this paper, we introduce a socially-attended attack to assess the social understanding of prediction models in terms of collision avoidance. An attack is a small yet carefully-crafted perturbations to fail predictors. Technically, we define collision as a failure mode of the output, and propose hard- and soft-attention mechanisms to guide our attack. Thanks to our attack, we shed light on the limitations of the current models in terms of their social understanding. We demonstrate the strengths of our method on the recent trajectory prediction models. Finally, we show that our attack can be employed to increase the social understanding of state-of-the-art models. The code is available online: https://s-attack.github.io/