Off The Beaten Sidewalk: Pedestrian Prediction In Shared Spaces For Autonomous Vehicles

TL;DR

This paper introduces a probabilistic, interpretable pedestrian prediction method for autonomous vehicles that performs well in diverse environments by modeling yielding behavior and vehicle influence.

Contribution

It proposes a novel risk-based attention mechanism and vehicle influence model to improve prediction accuracy and interpretability across various scene types.

Findings

Achieves state-of-the-art real-time prediction accuracy.

Performs well in both shared spaces and traditional scenes.

Provides interpretable predictions based on pedestrian yielding risk.

Abstract

Pedestrians and drivers interact closely in a wide range of environments. Autonomous vehicles (AVs) correspondingly face the need to predict pedestrians' future trajectories in these same environments. Traditional model-based prediction methods have been limited to making predictions in highly structured scenes with signalized intersections, marked crosswalks, or curbs. Deep learning methods have instead leveraged datasets to learn predictive features that generalize across scenes, at the cost of model interpretability. This paper aims to achieve both widely applicable and interpretable predictions by proposing a risk-based attention mechanism to learn when pedestrians yield, and a model of vehicle influence to learn how yielding affects motion. A novel probabilistic method, Off the Sidewalk Predictions (OSP), uses these to achieve accurate predictions in both shared spaces and traditional scenes. Experiments on urban datasets demonstrate that the realtime method achieves state-of-the-art performance.