Reinforcement Learning for Space-Time Access in MISO Uplink Underwater Acoustic Communications

Abstract

Underwater acoustic communication shows unique challenges arising from the complex and dynamic nature of the propagation medium. Traditional approaches to enhance the communication efficiency have often fallen short due to the inherent uncertainties and limitations of underwater channels. Recently, the advances in artificial intelligence offer promising to address these challenges. Specifically, this paper presents an overview of the potential application of reinforcement learning to handle multiple access in underwater acoustic networks exploiting space and time domains. By leveraging reinforcement learning algorithms, autonomous agents can adaptively learn optimal communication strategies in response to changing environmental conditions, improving channel usage and, therefore, data throughput, reliability, and energy efficiency.