Context & Justification
Landing is one of the most critical flight phases with the highest incident rate in autonomous UAV operations. In scientific missions conducted in remote areas of Ecuador (volcano edges, forest areas, post-disaster risk zones), the availability of prepared landing areas is minimal or nonexistent, requiring the development of autonomous landing capabilities on unstructured terrain.
This project develops a visual guidance system for precision landing that, through real-time image processing and the use of fiducial markers or AI-based terrain analysis, allows the UAV to identify a safe landing zone and execute the descent maneuver with a position error below 30 centimeters, without operator intervention.
Main Research Goal
Specific Scopes
- Design and implement a landing zone detection and tracking algorithm using image processing, combining ArUco marker detection and terrain homogeneity analysis via neural network.
- Develop a precision visual guidance controller integrating the estimated landing zone position with the flight control system (ArduCopter/PX4) via MAVLink protocol.
- Implement the system on low-weight embedded hardware (Raspberry Pi CM4 + RGB camera) and integrate it into a multirotor test UAV at ATA laboratory.
- Validate the system through at least 30 landing tests under different lighting, wind, and initial target distance conditions, quantifying the final positioning error.
Research Products
Embedded computer vision system for precision landing, with landing zone detection algorithms and visual guidance controller integrated and documented.
Autonomous landing test protocol with position error measurement methodology and statistical analysis of results under multiple operational conditions.
Experimental validation report with data from at least 30 test landings, positioning error distribution, and analysis of influencing factors.
Documented source code and system repository, available for integration into future ATA Research Group UAV platforms.