Super-Resolution of Wireless Channel Characteristics: A Multitask Learning Model
Xiping Wang, Ke Guan, Danping He, Zhao Zhang, Haoyang Zhang, Jianwu Dou, Zhangdui Zhong
Abstract
Channel modeling has always been the core part of the design and development of communication system, especially in the sixth-generation (6G) era. Traditional approaches like stochastic channel modeling and ray-tracing (RT)-based channel modeling depend heavily on measurement data or simulation, which are usually expensive and time-consuming. In this article, we propose a super-resolution (SR) model for recovering high-resolution (HR) channel characteristics from sparse sampling data. The model is based on multitask learning (MTL) convolutional neural networks (CNNs) with attention mechanism and residual connection. Experiments demonstrate that the proposed MTL SR model could achieve fairly good performances in terms of mean absolute error (MAE) and standard deviation of error (STDE). The advantages of the proposed model are demonstrated in comparison with other state-of-the-art deep learning (DL) models and visualization of SR performances. The ablation study also proved the necessity of data augmentation and the techniques in the model design. The good generalizability of the proposed MTL SR model leads to discussions of potential applications. The contribution of this article could be helpful in channel modeling, network optimization, positioning, and other wireless channel characteristics-related work by largely reducing the workload of simulation or measurement.