Autonomous underwater vehicle docking system for energy and data transmission in cabled ocean observatory networks

Abstract

Cabled ocean observatory networks (COON) are used for long-term all-weather observation of submarine scientific data, which contribute to low-carbon ocean energy research. Autonomous underwater vehicles (AUV) with clean energy can provide active search capabilities by connecting with the docking station (DS) on the COON to complete energy and data transmission in long-term detection tasks. The AUV is guided by optical active landmarks and a vision system for short-range docking. In this study, we propose an active landmarks tracking framework to solve the problem of detecting failure caused by incomplete observation of landmarks. First, a two-stage docking algorithm based on CNN is used to estimate the 3D relative position and orientation between DS and AUV during docking, including detect phase and PnP pose estimator. Then extended Kalman filter and Hungarian matching algorithm are introduced to improve the robustness of the algorithm. The reliability of the vision-based short-range docking algorithm is verified in the pool, and the robustness of the algorithm to the field environment is shown in the lake field experiment combined with long-range guidance. The experimental results indicate that the algorithm framework can effectively leverage the landmarks information and enhance the scope of the visual guidance algorithm.