ART: Adaptive Relational Transformer for Pedestrian Trajectory Prediction with Temporal-Aware Relations

TL;DR

This paper introduces ART, a novel transformer-based model with a temporal-aware relation graph and adaptive pruning, achieving state-of-the-art pedestrian trajectory prediction with high efficiency.

Contribution

The paper proposes ART, combining TARG and AIP mechanisms to better model dynamic human interactions and reduce computational costs.

Findings

ART achieves state-of-the-art accuracy on ETH/UCY and NBA benchmarks.

The model significantly reduces redundant computations compared to existing methods.

Experimental results demonstrate high efficiency and improved prediction performance.

Abstract

Accurate prediction of real-world pedestrian trajectories is crucial for a wide range of robot-related applications. Recent approaches typically adopt graph-based or transformer-based frameworks to model interactions. Despite their effectiveness, these methods either introduce unnecessary computational overhead or struggle to represent the diverse and time-varying characteristics of human interactions. In this work, we present an Adaptive Relational Transformer (ART), which introduces a Temporal-Aware Relation Graph (TARG) to explicitly capture the evolution of pairwise interactions and an Adaptive Interaction Pruning (AIP) mechanism to reduce redundant computations efficiently. Extensive evaluations on ETH/UCY and NBA benchmarks show that ART delivers state-of-the-art accuracy with high computational efficiency.