Vision driven trailer loading for autonomous surface vehicles in dynamic environments.

Automated docking technologies for marine vessels have advanced significantly, yet trailer loading, a critical and routine task for autonomous surface vehicles (ASVs), remains largely underexplored. This paper presents a novel, vision-based framework for autonomous trailer loading that operates without GPS, making it adaptable to dynamic and unstructured environments. The proposed method integrates real-time computer vision with a finite state machine (FSM) control strategy to detect, approach, and align the ASV with the trailer using visual cues such as LED panels and bunk boards. A realistic simulation environment, modeled after real-world conditions and incorporating wave disturbances, was developed to validate the approach and is available. Experimental results using the WAM-V 16 ASV in Gazebo demonstrated a 100% success rate under calm to medium wave disturbances and a 90% success rate under high wave conditions. These findings highlight the robustness and adaptability of the vision-driven system, offering a promising solution for fully autonomous trailer loading in GPS-denied scenarios.