A Reinforcement-Learning-Based Beam Adaptation for Underwater Optical Wireless Communications
Imene Romdhane, Georges Kaddoum
Abstract
Underwater optical wireless communications (UOWCs) have recently appeared as an attractive solution for many applications, such as remote controlling and sensing due to its advantages, such as high transmission rate, ultrawide bandwidth, and low latency. However, due to the harsh underwater conditions, UOWC faces challenges, such as water absorption, scattering, and pointing-acquisition-and-tracking (PAT) problems. This is mainly due to the dynamicity existing underwater. Consequently, this leads to packet loss and hence deteriorates the reliability and link quality of such networks. Such a problem can affect the degree of connectivity and end-to-end (E2E) performance of the communication system. The existing solutions in the literature are based on predefined models, assuming full knowledge of the environment. However, such models do not optimally treat the dynamicity existing underwater. This article proposes novel beam adaptation methods based on reinforcement learning (RL) for point-to-point UOWC. The first method aims to optimize the light beamwidth; the second method focuses on adapting the beam orientation, whereas the last one optimizes both the light’s beamwidth and beam orientation. Our proposed RL-based solutions yield optimal positioning and beamwidth of the light source and improve the considered communication link’s success rate. They also guarantee better link quality in terms of signal-to-noise ratio (SNR) compared to the uncertainty disk static method for four different underwater environments, including pure seawater, clean ocean, coastal ocean, and turbid harbor.