Deep Learning-Based Acquisition Pointing and Tracking for Underwater Wireless Optical Communication

Complex underwater environmental factors, such as turbidity, turbulence, background light, and mobility, pose great challenges to the rapid and stable establishment of underwater wireless optical communication (UWOC) links. Therefore, it is crucial to accelerate the development of fast, high-precision, and robust acquisition, pointing, and tracking (APT) systems to promote the practical application of UWOC. In this work, the YOLOv5s algorithm with strong learning and adaptive abilities and a solar panel receiver with a large detection area are proposed to solve the link alignment issues caused by complex underwater environmental factors. In a 3-m water tank, the turbidity, bubbles-induced turbulence, and background light have less impact on the robust YOLOv5s-based APT system. The average pointing time is merely 5.97 ms, and the average tracking time is merely 229 ms for different underwater environments. When the concentration of Mg(OH)2 reaches 3.39 mg/L and the scintillation index of bubbles reaches 0.019, the UWOC link can still be established quickly with a data rate of 1.76 Mbit/s, respectively. Moreover, the APT system can tolerate the illuminance of the background light as low as 0.2 lx. We also conduct a preliminary study on the performance of the system in mobile conditions, which proves the effectiveness of the system and lays a foundation for further research.