Semi-Autonomous Aircraft Tug Docking Using a Stereo Camera

Sammanfattning

Towbarless aircraft tugs dock by lifting the aircraft by placing the airplanes nose gear in acradle and lifting it up, handing over the control to the tug. This docking procedure requiresa skilled driver, and driver assistance functions are desired to improve the efficiency. Thisthesis investigates the development of a semi-autonomous driver assistance system for thedocking procedure.The proposed perception and navigation pipeline utilizes a ZED X stereo camera withina ROS2 framework. A custom-trained YOLO model is employed to detect and localize thenose gear, while an Iterative Closest Point (ICP) algorithm, combined with bounding boxdata, estimates the yaw angle from the 3D point cloud. For autonomous vehicle control, apath-following navigation system based on Dubins paths and a Pure Pursuit algorithm wasimplemented. The system relies on the camera’s built-in VSLAM for odometry and transmitscontrol signals via a CAN bus interface.Experimental evaluations demonstrate that while the YOLO model successfully localizesthe nose gear, its robustness is highly sensitive to environmental lighting and weather conditions.Furthermore, yaw angle estimation is only accurate at short ranges and narrow approachangles. Consequently, the limited effective range of the stereo camera prevents timelytrajectory corrections, given the tug’s large turning radius. Additionally, the VSLAM systemexhibits significant drift, causing the Pure Pursuit controller to overshoot the target pose.In conclusion, while the developed software architecture and algorithms demonstrateclear potential for autonomous docking, the current hardware limits system reliability. Achievingthe required operational precision necessitates upgrading the sensor suite to support accuratelong-range spatial estimation.