ART: Adaptive Relational Transformer for Pedestrian Trajectory Prediction with Temporal-Aware Relations
This work addresses the problem of predicting pedestrian trajectories for robot-related applications, offering an incremental improvement over existing graph-based or transformer-based methods.
The paper tackles pedestrian trajectory prediction by proposing an Adaptive Relational Transformer (ART) with a Temporal-Aware Relation Graph and Adaptive Interaction Pruning, achieving state-of-the-art accuracy and high computational efficiency on ETH/UCY and NBA benchmarks.
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.