Reinforcement Learning-Based Energy-Efficient Power Allocation For Underwater Full-Duplex Relay Network With Energy Harvesting

In this paper, we study the energy efficiency (EE) performance of a three-node underwater full-duplex relay network, where the relay is an energy harvesting node. Since the arrival of harvested energy is intermittent from the ambient environment, energy-efficient data transmission can prolong the lifespan of the network. By exploiting the causal system information, we aim to maximize the long-term end-to-end EE of the network through adaptive power control at the relay node. The system is described through a Markov decision process, and the reinforcement learning framework is applied to obtain the energy-efficient transmission policy. Simulation results demonstrate the long-term average EE performance of the obtained transmission policy, which outperforms two benchmark approaches.