Adaptive Modulation Scheme for Satellite Communication Channel Based on RLNN

Abstract

This paper discusses the potential applications of reinforcement learning in communication systems. In order to improve the spectral efficiency of wireless communication channels, we propose an adaptive modulation scheme based on reinforcement learning and deep neural network (RLNN) with a average network exploration(AE) strategy. Although the application of machine learning technology in link adaptation has attracted widespread attention, most of the solutions currently proposed are based on offline training algorithms, which are not suitable for real-time operations. Classical reinforcement learning is only effective in discrete actions and states. The proposed technical solution does not depend on the offline training mode. The proposed technical solution uses AE strategy to drive exploration, which can improve the exploration efficiency of the agent. The simulation result shows that we can increase the data transmission rate under limited bandwidth and independently improves link availability, throughput and the efficiency of the wireless communication system.