Article Overview
This paper proposes a perception-guided framework for autonomous UAV landing in cluttered, unstructured environments. RGB-D images and point clouds are processed to characterize surface flatness, steepness, inclination, and depth variation. A fuzzy inference system combines these terrain metrics into a landing suitability map.
Visual simultaneous localization and mapping (vSLAM) estimates the UAV’s pose and generates sparse 3D map points. These points are projected onto the image plane and filtered using a thresholded landing suitability map. Candidate regions are assessed using clustering, surface geometry, and the area required for landing to select a suitable location.
Four synthetic corner points are then defined around the selected location based on the UAV’s dimensions. Their projected image features are used for image-based visual servoing (IBVS) during landing. The complete approach is evaluated in ROS–Gazebo simulations and compared with an existing landing-site detection method.
Key Features
- Fuzzy landing suitability map: Combines depth and surface-geometry information to identify potentially safe landing regions.
- vSLAM-based landing-site selection: Filters mapped 3D points and evaluates candidate regions using geometric and area constraints.
- Image-based landing control: Uses synthetic corner points and image moments to guide the UAV toward the selected landing site.
Figures
Proposed landing pipeline: sensory inputs (green), vSLAM (yellow), fuzzy landing map (red), and IBVS control (purple).
Read More
Read the full paper.
Citation
@article{sepahvand2024novel,
title={A Novel Fuzzy Image-Based UAV Landing Using RGBD Data and Visual SLAM},
author={Sepahvand, Shayan and Amiri, Niloufar and Masnavi, Houman and Mantegh, Iraj and Janabi-Sharifi, Farrokh},
journal={Drones},
volume={8},
number={10},
pages={594},
year={2024},
publisher={MDPI}
}