”The project successfully developed a demonstrable solution for autonomous drone navigation that identifies individual trees using image recognition and centers the drone with high precision for targeted fertilization in complex forest environments.”, says Per Bröms, Innovation Lead, Aero EDIH.
Challenges
Nordluft, in collaboration with SLU (Swedish University of Agricultural Sciences), identified a significant opportunity to improve forest management through drone technology. The concept focused on dose fertilization of young trees under 3 meters in height, which require precise nutrient delivery above the stem to ensure proper uptake. This approach offers economic benefits, reduces nitrogen leakage compared to conventional methods, and improves tree survival rates.
To realize this concept, Nordluft needed a robust solution for autonomous drone navigation close to the ground in forest environments. Such navigation is highly challenging due to uneven terrain, dense vegetation, and obstacles. Solving this problem would not only enable precision fertilization but also open doors to other applications such as seedling transport, wildlife protection treatments, and health monitoring.
Solutions
Aero EDIH supported Nordluft through Explore Autonomy AB with consultants Casper and Koray. The project began with preparatory meetings and sensor readiness checks, followed by testing at Barkarby airfield and in a forest environment north of Uppsala. The test site featured a mix of clear-cut areas, scattered mature trees, and dense mats of small trees (1–3 meters high).
The delivered solution enables autonomous navigation between predefined GPS waypoints representing target trees. Upon reaching a waypoint, the drone activates a stereo camera connected to an NVIDIA Jetson module (an embedded computing platform) running an AI-based image recognition system to identify the tree. This compensates for GPS inaccuracies and ensures precise positioning above the stem.
Once the tree is identified, the system sends correction data to the autopilot, which centers the drone and lowers it to approximately 1 meter above the tree top, avoiding obstacles detected by the stereo camera and forward-facing radar. The fertilization mechanism then releases a dose of nutrients before the drone ascends and moves to the next target.
Results and Benefits
The project delivered a demonstrable solution for autonomous close-to-ground navigation in relevant forest conditions. This capability enables precision fertilization, reducing waste and environmental impact while improving growth and survival rates of young trees.
Operationally, the system combines GPS navigation with AI-driven visual recognition, ensuring accurate positioning even in challenging environments. The integration of radar and stereo vision enhances safety by preventing collisions with surrounding vegetation.
Strategically, this solution positions Nordluft as a pioneer in precision forestry, with potential to expand into related applications such as automated seedling planting and health monitoring. These advancements support sustainable forestry practices and contribute to digital transformation in rural industries.
Perceived Social and Economic Impact
Autonomous precision fertilization reduces chemical waste, minimizes environmental risks, and improves forest productivity. By enabling efficient and targeted operations, this technology supports sustainable forestry, lowers operational costs, and enhances ecosystem health. The approach also creates opportunities for innovation in other forestry applications, contributing to economic growth and environmental stewardship.
Lessons Learned
Combine multiple sensing modalities (GPS, stereo vision, radar) to achieve robust navigation in complex environments. Validate AI models under real-world conditions and integrate correction loops for precise positioning.
Do not rely solely on GPS for close-to-ground operations. Positional errors can compromise accuracy and safety. Avoid static waypoint strategies for long-term scalability – dynamic area scanning and adaptive tree detection are essential for future improvements.
“The support from Aero EDIH meant that Nordluft could take the first steps in the development of a family of new very exciting forestry applications for drones. All of these depend on autonomous efficient navigation close to the ground in forest environments, with very challenging terrain. The project resulted in a successful demonstration. The base is now in place for further development and commercialization of, among other things, dosed fertilization of small trees. An action with very large economic potential for forestry where the environmental impact is kept minimal”, says Elof Winroth, Founder, Nordluft Automation.
