SCAN: A Spatial Context Attentive Network for Joint Multi-Agent Intent Prediction

TL;DR

SCAN is a novel neural network that predicts multiple socially-acceptable future pedestrian trajectories by effectively modeling spatial influences with a new attention mechanism, outperforming existing methods.

Contribution

Introduces a spatial attention mechanism within a joint prediction framework that is more efficient and interpretable than previous approaches.

Findings

Outperforms state-of-the-art trajectory prediction methods.

Efficiently encodes spatial influence with fewer assumptions.

Demonstrates superior accuracy across multiple datasets.

Abstract

Safe navigation of autonomous agents in human centric environments requires the ability to understand and predict motion of neighboring pedestrians. However, predicting pedestrian intent is a complex problem. Pedestrian motion is governed by complex social navigation norms, is dependent on neighbors' trajectories, and is multimodal in nature. In this work, we propose SCAN, a Spatial Context Attentive Network that can jointly predict socially-acceptable multiple future trajectories for all pedestrians in a scene. SCAN encodes the influence of spatially close neighbors using a novel spatial attention mechanism in a manner that relies on fewer assumptions, is parameter efficient, and is more interpretable compared to state-of-the-art spatial attention approaches. Through experiments on several datasets we demonstrate that our approach can also quantitatively outperform state of the art trajectory prediction methods in terms of accuracy of predicted intent.