Towards an intelligent reliable connectivity for underwater wireless communication

Abstract

Nowadays, the underwater environment is receiving increasing attention for deploying communications in different fields. Thus, finding a robust mean of underwater communication against the harsh underwater environment conditions is becoming crucial. Acoustic and radio frequency (RF) communications were initially deployed for underwater communications. However, acoustic communications suffer from high latency and limited beamwidth, and RF communications suffer from high energy consumption and severe attenuation underwater. Consequently, underwater optical wireless communications (UOWC) were lately proposed as a promising candidate. Thanks to the remarkable advantages of optical waves, such as the high speed of light, high data rates, and low power consumption, it became a center of interest for many researches. Yet, UOWC suffer from limitations that can degrade the communication quality, such as short communication ranges and pointing errors (PEs). The main objective of this thesis is to construct an autonomous self-organized network (SON) for UOWC that is robust against the harsh underwater environment. The first goal is to maintain the connectivity between the nodes by solving the beam’s alignment problem. The second goal is to extend the network’s lifetime by optimizing the nodes’ power consumption. The third goal is to solve the fairness problem in UOWC networks. In this vein, we first propose reinforcement learning (RL)-based beam adaptation solutions for UOWC point-to-point (P2P) communication. These solutions optimize the light beamwidth, beam orientation, or both. These solutions improve the optical link’s success rate and guarantee better link quality in terms of signal to-noise ratio (SNR) for four different underwater environments, including pure seawater, clean ocean, coastal ocean, and turbid harbor. Then, we analyse the outage probability for UOWC over oceanic turbulence with PEs. A combined UOWC system including a maximum ratio combining (MRC)-based direct link and an optical intelligent reflecting surface (OIRS)-assisted indirect link is considered. The exponential-generalized gamma (EGG) distribution is adopted to model the oceanic turbulence. The proposed system proves to enhance the connectivity in high-speed UOWC under different underwater channel conditions. Furthermore, we consider a direct air-to-underwater P2P communication where a terrestrial unmanned aerial vehicle (UAV) is communicating with an underwater node directly with no relay on the water surface. Specifically, we present a deep deterministic policy gradient (DDPG)- based link adaptation technique, considering the orthogonal frequency division multiplexing (OFDM) technique under noisy channel state information (CSI) conditions. This technique optimizes the beamwidth, beam orientation, and transmitting power levels from the UAV, treating these parameters as continuous variables to account for the inherent randomness in real-world scenarios. The proposed method proves to maintain high success rates, while reducing the energy consumption compared to Q-learning based approach. Finally, we introduce dynamic DDPG-based beamforming optimization solutions for multi-beam UOWC networks. Non-orthogonal multiple access (NOMA) is adopted within each beam to support multi user communications. This technique optimizes the beam orientation and power allocation among different receiving nodes, considering node’s quasi-stationarity due to underwater dynamicity. Both a single-agent and a sequential multi-agent DDPG-based solutions are proposed in this chapter, in the aim of maximizing the system’s energy efficiency (EE) while guaranteeing the max-min fairness.

Date	28 Jun 2026
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Georges Kaddoum (Supervisor)