Robust Monocular Localization of Drones by Adapting Domain Maps to Depth Prediction Inaccuracies

TL;DR

This paper introduces a monocular drone localization method that combines deep learning-based depth prediction with Bayesian filtering, achieving high accuracy and robustness in challenging environments without extensive domain adaptation.

Contribution

It proposes a cross-modal framework that jointly trains depth prediction and pose reasoning, improving scalability and environmental robustness over deep learning-only methods.

Findings

Maintains pose accuracy with poor depth estimates from lightweight predictors.

Performs well under extreme lighting conditions without explicit domain adaptation.

Enables faster updates and resource-efficient reuse of intermediate predictions.

Abstract

We present a novel monocular localization framework by jointly training deep learning-based depth prediction and Bayesian filtering-based pose reasoning. The proposed cross-modal framework significantly outperforms deep learning-only predictions with respect to model scalability and tolerance to environmental variations. Specifically, we show little-to-no degradation of pose accuracy even with extremely poor depth estimates from a lightweight depth predictor. Our framework also maintains high pose accuracy in extreme lighting variations compared to standard deep learning, even without explicit domain adaptation. By openly representing the map and intermediate feature maps (such as depth estimates), our framework also allows for faster updates and reusing intermediate predictions for other tasks, such as obstacle avoidance, resulting in much higher resource efficiency.