Intelligent Adaptive MIMO Transmission for Nonstationary Communication Environment: A Deep Reinforcement Learning Approach
Xin Lin, Aijun Liu, Chen Han, Xiaohu Liang, Yifu Sun, Guoru Ding, Haibo Zhou
Abstract
Multiple-input multiple-output (MIMO) technology can effectively improve transmission throughput and reliability by utilizing spatial wireless resources, which has aroused widespread research attentions. Comparing with the stationary communication environment considered in most studies, the nonstationary channel may cause severe performance degradation of MIMO technology. This promotes the research of adaptive MIMO transmission strategy, which can intelligently adapt to dynamic environment and provide reliable and efficient communication. In this paper, our purpose is to design an intelligent MIMO system that can adjust the MIMO transmission mode and modulation order according to nonstationary environment. The dynamic decision problem is formulated as a markov decision process (MDP) and the state, action and reward function are designed. Then, an adaptive MIMO transmission strategy via leveraging proximal policy optimization (PPO) learning framework is proposed. The trained PPO agent can learn the proper joint transmission strategy so as to maximize the spectral efficiency subject to the constraint of target bit error rate (BER) performance. Simulation results demonstrate that the proposed scheme can achieve significant performance improvement over benchmark schemes.