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

arXiv:2405.09049v1 Announce Type: new
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 …

abstract active learning arxiv autonomous autonomous driving clustering cs.ai cs.cv cs.lg cs.ro curation data data curation driving dynamic dynamics framework information machine machine learning perception prediction representation sampling state strategies study tasks trajectory type vision