•  
  •  
 

Abstract

Traditionally, high-value fish have been located at sea using helicopters. However, with advancements in technology, maritime drones have become increasingly important in recent years. Compared to traditional helicopter-based methods, drones offer significantly lower operational costs, making them cost-effective. However, the challenging sea conditions, including strong winds and the swaying motion of vessels, pose significant challenges for drone landings.

This paper proposes a stable approach for multicopter landing on ships at sea, incorporating improved marker detection, enhanced wind resistance control, and more accurate deck motion prediction. In the marker detection phase, we introduce a method to improve the accuracy of ArUco marker contour and corner detection in low-resolution images. For wind resistance control, we use the positional error between the world coordinate system and the ArUco marker as input and regulate the drone’s motion using a PID-controlled velocity component. To predict deck motion, we employ an online Backpropagation Neural Network to estimate the ship’s roll angle and determine the optimal landing timing.

Our approach is evaluated through land-based simulation experiments and compared with existing methods. Most importantly, we validate its effectiveness through real-world landing tests in an actual maritime environment. The landing error was consistent with those observed in land-based experiments, with an average X and Y error of 5.731 cm, indicating the robustness of the system even in offshore conditions. In comparison to existing drone landing research, this study presents significant improvements, particularly with the precision landing system, offering an 18.13% improvement over the best-known method, which has a 7 cm error. This demonstrates that our system can achieve a significantly higher accuracy, with practical implications for offshore drone operations.

Download
COinS