Statistically validated comparative evaluation of fiducial markers for vision-based UAV precision landing in standardized PX4–ROS simulation

Abstract

UAV quadcopter technology has advanced rapidly and is widely used in both military and civilian sectors. One critical challenge is precision landing, since conventional GPS-based methods have limited accuracy (0.9–1.8 m). Fiducial marker–based landing offers a practical alternative, but cross-study comparisons remain difficult because marker types and testing procedures vary. This study presents a statistically validated comparative evaluation of five fiducial markers (ArUco, AprilTag, ARTag, WhyCode, and STag) in a standardized PX4–ROS simulation environment. Beyond performance ranking, the study analyzes how marker geometry and coding structure influence homography stability, pose-estimation noise, and landing control response. Evaluation metrics include horizontal and vertical landing errors, landing time, and detection distance, analyzed using one-way ANOVA and Tukey HSD tests. Results show significant differences in vertical error, landing time, and detection distance, indicating that geometric regularity and edge-contrast structure directly affect visual pose stability. Within the controlled simulation framework, AprilTag shows the most consistent overall performance, with mean errors of 2.99 cm (X-axis) and 5.59 cm (Y-axis) and an average landing time of 21.8 s. The main contribution is a reproducible comparative framework and an explanatory link between marker design and visual-control stability, guiding marker selection in precision UAV landing systems.