Perception Without Vision for Trajectory Prediction: Ego Vehicle Dynamics as Scene Representation for Efficient Active Learning in Autonomous Driving

TL;DR

This paper presents a novel active learning approach for autonomous driving that uses trajectory and dynamic state information to select training data efficiently, reducing annotation costs while maintaining high model performance.

Contribution

It introduces trajectory-based clustering and sampling strategies within an active learning framework, improving data efficiency for autonomous vehicle trajectory prediction.

Findings

Achieves sub-baseline displacement errors with only 50% of data

Demonstrates consistent performance gains over random sampling

Effective in reducing data annotation costs

Abstract

This study investigates the use of trajectory and dynamic state information for efficient data curation in autonomous driving machine learning tasks. We propose methods for clustering trajectory-states and sampling strategies in an active learning framework, aiming to reduce annotation and data costs while maintaining model performance. Our approach leverages trajectory information to guide data selection, promoting diversity in the training data. We demonstrate the effectiveness of our methods on the trajectory prediction task using the nuScenes dataset, showing consistent performance gains over random sampling across different data pool sizes, and even reaching sub-baseline displacement errors at just 50% of the data cost. Our results suggest that sampling typical data initially helps overcome the ''cold start problem,'' while introducing novelty becomes more beneficial as the training pool size increases. By integrating trajectory-state-informed active learning, we demonstrate that more efficient and robust autonomous driving systems are possible and practical using low-cost data curation strategies.