Achieving Time-Sharing and Spatial-Reuse Underwater Wireless Sensor Networks with Communication Fairness: A Distributed Deep Multi-Agent Reinforcement Learning Approach

Abstract

It is difficult to provide qualified and fair communications for time-sharing and spatial reuse underwater wireless sensor networks when energy supplements are limited, the environment is non-stationary, and communication interference is strong. Due to the physical separation of underwater nodes, transmissions are intended to occur concurrently to maximize network capacity. Currently available approaches for improving fairness are frequently at the expense of network capacity. Methods that seek a better trade-off between fairness and network capacity are required. This paper proposes a novel approach to maximize network capacity and improve communication fairness by increasing simultaneous communications, achieving time-sharing, and spatial-reuse UWSNs. It is a distributed, multi-agent reinforcement learning approach that utilizes an observation encoder and a local utility network to coordinate collaboration across underwater nodes by adaptively tuning transmit parameters in response to local observations. In terms of network capacity, fairness, and reuse, we compared the suggested methodology to standard methods. Experiments reveal that, when compared to other ways, ours maximizes reuse and produces a significantly superior trade-off between network capacity and fairness, while still meeting lifetime and energy restrictions. The work presented in this article is anticipated to develop into valuable tools for designing and optimizing UWSNs.