Semi-Deterministic Dynamic Millimeter-wave Channel Modeling Based on an Optimal Neural Network Approach
Xiongwen Zhao,Zihao Fu,Wei Fan,Yu Zhang,Suiyan Geng,Fei Du,Peng Qin,Zhenyu Zhou,Lei Zhang
DOI: https://doi.org/10.1109/tap.2022.3145438
IF: 5.7
2022-01-01
IEEE Transactions on Antennas and Propagation
Abstract:Billions of mobile terminals will be deployed in various Internet of Things (IoT), in which millimeter-wave (mmWave) technology will be gradually applied. Accurate modeling and simulation of wireless channel is the base for efficient design and performance evaluation. This becomes more important for industrial scenarios, which might be highly dynamic and potentially different from well-investigated cellular deployment scenarios. In this work, a novel semi-deterministic mmWave dynamic channel modeling approach based on optimal neural network (ONN) principle is proposed. The ONNs are radial basis function (RBF) neural networks (NNs) trained with optimal variance parameters and are applied to predict large-scale channel parameters (LSCPs) [e.g., path loss (PL), delay spread (DS), and angle spread (AS)]. Based on the ONNs' predicted large-scale parameters and simplified propagation environment including the layout of transmitter (TX), receiver (RX), and major scatterers, the proposed channel modeling approach can generate accurate dynamic channel parameters. The proposed approach is validated by the channel data measured at a high-voltage substation. Large-scale parameters, multipath component (MPC) distributions, and power delay profiles (PDPs) are validated. The proposed approach is demonstrated to be an accurate, fast, and robust channel modeling method, which can be used for both link-level and system-level channel simulation for future design and optimization of industrial IoT.
telecommunications,engineering, electrical & electronic