Intelligent Propagation Prediction Model for Wireless Radio Channel Based on CNN

Abstract

AbstractAs a reliable communication medium, the wireless radio channel determines the quality and performance of wireless communication systems. It is imperative for operators to accurately predict the propagation model before designing and planning the next-generation cellular networks. The intelligent propagation model of wireless radio channels based on the convolutional neural network (CNN) is proposed to predict reference signal receiving power (RSRP) in this paper. Specifically, several appropriate features are designed in terms of the transmitter, propagation, and receiver according to the empirical models, such as the Okumura-Hata model and the Cost 231-Hata model. And then these features with high variance and correlation coefficient are selected. Data preprocessing is carried out through data cleaning, data normalization, data partition, and data visualization before the training stage of the intelligent propagation model. Finally, the selected features are fed into CNN to predict the RSRP value. The simulation results indicate that the proposed model has superior performance. The work of this paper will contribute to the planning and optimization of future cellular networks.KeywordsIntelligent propagation prediction modelWireless radio channelConvolutional Neural Network (CNN)RSRP prediction