Adaptive Transmission for Underwater Acoustic Communication Based on Deep Reinforcement Learning

Due to the complexity of the underwater environment and the harsh channel conditions, energy-efficient transmission has become a key topic in underwater acoustic (UWA) communication. In this paper, we propose a UWA adaptive communication strategy that aims to maximize the long-term energy efficiency of a single link. To cope with the rapid changes in the UWA channel, we employ the deep reinforcement learning (DRL) method to find the optimal transmission policy by jointly scheduling the transmit power, modulation order, and coding rate. The proposed method first learns the channel variation law. Then, the feedforward channel is predicted based on the channel state of the feedback link, and appropriate transmission parameters are selected to achieve a trade-off between throughput and power consumption. Simulation results show that the adaptive communication scheme is superior to the fixed modulation and coding scheme (MCS) when the channel is known, and can greatly improve the system’s energy efficiency.