Autonomous drone system with integrated vision-based control for marine buoy retrieval

Department of Mechatronics, Nelson Mandela University, Gqeberha, 6013, South Africa

^* Corresponding author: stefan.vanaardt@mandela.ac.za

Abstract

Retrieving marine drifter buoys is often resource-intensive, requiring vessels and personnel to locate and recover floating sensors manually. This paper presents a fully autonomous quadrotor drone system designed to retrieve marine buoys using computer vision and a custom lightweight gripper. An ArUco marker affixed to the buoy enables visual pose estimation from a downward-facing camera. A gripper that doubles as landing gear minimises the payload weight. The control architecture integrates a Pixhawk flight controller for low-level stability and a Raspberry Pi running ROS for vision processing and mission logic. In indoor trials, the system achieved an 80% success rate in marker detection and 95% manual grasping success. Precision descent resulted in a 53 mm average error. Integration of the gripper into the landing structure increased flight time from 8.7 to 12.5 minutes. This research demonstrates the feasibility of autonomous aerial retrieval and identifies enhancements required for open-water deployment.