Adaptive UAV Visual Localisation Based on Improved Gradient-Damping Newton Method
Open Access
- 7 May 2026
- journal article
- research article
- Published by MDPI AG in Electronics
- Vol. 15 (10) , 1974
Abstract
The role of unmanned aerial vehicles (UAVs) in time-sensitive missions such as low-altitude reconnaissance and disaster rescue has gained increasing significance. To address the challenge of visual localisation for UAVs operating in complex terrains under Global Navigation Satellite System (GNSS)-denied environments, this paper proposes an improved adaptive gradient-damped Newton approach to mitigate the trade-off between terrain non-convexity and computational real-time performance. The proposed approach incorporates a terrain-gradient-based dynamic step-size adjustment mechanism that adaptively captures non-linear terrain characteristics in real time and effectively reduces the numerical oscillations typically observed in steep regions when using the standard Newton method. In addition, a tightly coupled vision–geometry framework was developed to constrain cumulative drift during long-range flight. Monte Carlo simulation results demonstrate that the proposed algorithm maintains submeter localisation accuracy while achieving approximately a three-fold improvement in computational efficiency compared with traditional grid-based methods, and a 27.4% increase in convergence speed relative to the standard Newton method. Experiments conducted under high-noise conditions and highly undulating terrains indicate that the approach exhibits strong convergence stability, offering a computationally efficient and robust solution for UAV navigation.Keywords
Funding Information
- Natural Science Foundation of Sichuan province, China (24NSFSC2722)
- National Natural Science Foundation of China (62406043)
This publication has 34 references indexed in Scilit:
- A Review of UAV Positioning in LAIN: Toward a 5G-Core “Space-Air-Ground” Integrated and Cooperative ArchitectureENGINEERING Information Technology & Electronic Engineering, 2026
- OKVIS2-X: Open Keyframe-Based Visual-Inertial SLAM Configurable With Dense Depth or LiDAR, and GNSSIEEE Transactions on Robotics, 2025
- SED-UAV: A Synergistic Framework of Lightweight Chaotic Encryption and Multiscale Feature Detection for Secure UAV ApplicationsIEEE Internet of Things Journal, 2025
- Gnss-denied unmanned aerial vehicle navigation: analyzing computational complexity, sensor fusion, and localization methodologiesSatellite Navigation, 2025
- Deep learning-based 6-DoF visual relocalization assisted simultaneous localization and mapping (SLAM)Industrial Robot: the international journal of robotics research and application, 2025
- LWUAVDet: A Lightweight UAV Object Detection Network on Edge DevicesIEEE Internet of Things Journal, 2024
- Deep Learning for Visual Localization and Mapping: A SurveyIEEE Transactions on Neural Networks and Learning Systems, 2023
- Simultaneous Localization and Mapping: Through the Lens of Nonlinear OptimizationIEEE Robotics and Automation Letters, 2022
- Topographic SLAM Using a Single Terrain Altimeter in GNSS-Restricted EnvironmentIEEE Access, 2022
- UAV-Satellite View Synthesis for Cross-View Geo-LocalizationIEEE Transactions on Circuits and Systems for Video Technology, 2021