Autonomous docking is the key to the cooperative operation of unmanned underwater vehicles (UUVs). However, due to environmental complexity and object characteristics, it is very difficult to achieve precise guidance and docking. In order to improve the accuracy and robustness of underwater docking, this study proposes a vision-guided docking scheme which encompasses vision processing and 3D trajectory tracking control.

First, the overall vision-guided docking scheme is designed in combination with an analysis of task and object characteristics. Second, the YOLOv5 neural network is designed to complete the target detection of the underwater docking station, and the online measurement of the relative position and attitude relationship between the docking station and UUV is realized by an efficient perspective-n-point (EPnP) algorithm. Next, combined with the visual measurement results, an effective 3D robust trajectory tracking controller is designed on the basis of the 3D LOS guidance law, radial basis function neural network (RBFNN) and terminal sliding mode control (TSMC). Finally, the validity of the proposed scheme is verified through numerical simulation and a tank test.

In the tank test, the proposed vision-guided control algorithm can effectively complete the online detection and relative positioning of the underwater docking station, thereby achieving precise underwater docking.

The results of this study show that the proposed vision-guided 3D trajectory tracking control scheme is reasonable and efficient, and can lay a good foundation for UUV docking.