LARD 2.0: Enhanced Datasets and Benchmarking for Autonomous Landing Systems

arXiv cs.RO / 3/31/2026

📰 NewsSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces LARD 2.0 to address limitations in datasets used for supervised machine learning training for object detection in autonomous landing systems.
It expands dataset diversity by recommending new input sources—such as BingMap aerial imagery and Microsoft Flight Simulator—fed through an existing LARD dataset generator.
It refines the Operational Design Domain (ODD) by correcting unrealistic landing scenarios and extending coverage to multi-runway airports.
The authors propose a benchmarking framework to evaluate object-detection performance in a complex multi-instance setting and release open-source baseline models.
Overall, the work targets more representative training data and more reliable evaluation for ML components that support autonomous landing.

Abstract

This paper addresses key challenges in the development of autonomous landing systems, focusing on dataset limitations for supervised training of Machine Learning (ML) models for object detection. Our main contributions include: (1) Enhancing dataset diversity, by advocating for the inclusion of new sources such as BingMap aerial images and Flight Simulator, to widen the generation scope of an existing dataset generator used to produce the dataset LARD; (2) Refining the Operational Design Domain (ODD), addressing issues like unrealistic landing scenarios and expanding coverage to multi-runway airports; (3) Benchmarking ML models for autonomous landing systems, introducing a framework for evaluating object detection subtask in a complex multi-instances setting, and providing associated open-source models as a baseline for AI models' performance.