Researchers from the University of Nevada, Las Vegas, led by Ashik E Rasul, have developed a lightweight system for autonomous landing of QuadPlanes, a type of unmanned aerial vehicle that combines the long-range efficiency of fixed-wing aircraft with the maneuverability of multi-rotor drones. This research, published in the journal IEEE Access, focuses on enabling reliable autonomous landings in challenging environments, which is crucial for applications such as aerial monitoring and inspection in the energy sector.
QuadPlanes are increasingly being used for long-range autonomous missions, particularly in the energy industry for tasks like inspecting power lines, pipelines, and wind turbines. However, landing these aircraft autonomously in unstructured or GPS-denied environments presents significant challenges. The researchers addressed this by developing a perception-based landing system that uses deep neural networks to identify suitable landing sites across diverse visual and environmental conditions.
One of the key challenges in real-world deployment is the limited payload and volume available for onboard computing. The researchers optimized their system to work with edge AI devices like the NVIDIA Jetson Orin Nano, which are capable of real-time detection and control despite their compact size. They also developed a visual-inertial odometry system for accurate pose estimation during descent, which is essential for stable landings in the absence of GPS.
The flight characteristics of large QuadPlanes, such as high inertia and slow response times, further complicate autonomous landing maneuvers. The researchers designed their system to account for these factors, ensuring reliable operation even with the unique dynamics of these aircraft. The hardware platform, sensor configuration, and embedded computing architecture were all carefully optimized to meet the demanding real-time constraints of autonomous landing.
This research establishes a foundation for deploying autonomous landing systems in dynamic, unstructured, and GPS-denied environments. For the energy sector, this technology could enhance the safety and efficiency of aerial monitoring and inspection tasks, reducing the need for human intervention in potentially hazardous environments. As the use of unmanned aerial vehicles continues to grow in the energy industry, advancements in autonomous landing capabilities will be crucial for expanding their operational range and reliability.
This article is based on research available at arXiv.