Online Residual Learning from Offline Experts for Pedestrian Tracking

TL;DR

This paper introduces Online Residual Learning (ORL), a novel method combining offline predictions with online residual learning and expert advice to improve pedestrian trajectory prediction.

Contribution

The paper proposes ORL, integrating offline-trained predictors with online residual learning and expert advice, providing theoretical guarantees and improved pedestrian tracking performance.

Findings

ORL achieves best-of-both-worlds performance on Stanford Drone Dataset.

The method demonstrates effective online adaptation of offline predictions.

Guarantees are provided for ORL's regret bounds.

Abstract

In this paper, we consider the problem of predicting unknown targets from data. We propose Online Residual Learning (ORL), a method that combines online adaptation with offline-trained predictions. At a lower level, we employ multiple offline predictions generated before or at the beginning of the prediction horizon. We augment every offline prediction by learning their respective residual error concerning the true target state online, using the recursive least squares algorithm. At a higher level, we treat the augmented lower-level predictors as experts, adopting the Prediction with Expert Advice framework. We utilize an adaptive softmax weighting scheme to form an aggregate prediction and provide guarantees for ORL in terms of regret. We employ ORL to boost performance in the setting of online pedestrian trajectory prediction. Based on data from the Stanford Drone Dataset, we show that ORL can demonstrate best-of-both-worlds performance.