학술논문
NPSR: Neural Network enabled Phase-Space Reconstruction for Wireless Channel Prediction
Document Type
Conference
Author
Source
2022 IEEE Globecom Workshops (GC Wkshps) Globecom Workshops (GC Wkshps), 2022 IEEE. :1729-1735 Dec, 2022
Subject
Language
Abstract
In the massive multiple input output (MIMO) scenario of 5G wireless communication, the channel aging problem in high-speed mobile scenarios or rapidly changing channel propagation environment will significantly affect the performance of the communication system. Based on chaos theory, this paper proposes neural network enabled phase-space reconstruction(NPSR) to predict the channel state in MIMO, which alleviates the influence of channel aging. First, we show that the channel sequence is chaotic and obtain the reconstruction parameters of delay time and embedding dimension. Then, we combine the reconstructed phase-space and the back propagation (BP) network to predict the MIMO channel variation under different user moving speeds. We prove that the channel sequence is chaotic by using the Lyapunov exponent. In high-speed scenarios, NPSR outperforms the empirical chaos algorithm (ECA) and the autoregressive moving average (ARMA) algorithm. Compared with ARMA algorithms the average relative error of the prediction accuracy is reduced by 8.15%. Finally, we design an adjustable parameters configuration method based on ray features. The average time cost of NPSR in calculating phase-space reconstruction parameters is 37.6% of that of ECA.