Predictive Generation for Digital Twin Channel:Long-Batch Time-Frequency Construction  

作　　者：Che Chang Hu Jie Kang Honghui Rui Hua Lyu Xingzai Wang Bo 

机构地区：[1]School of Information and Communication Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China [2]State Key Laboratory of Mobile Network and Mobile Multimedia Technology,ZTE Corporation,Shenzhen 518055,China

出　　处：《China Communications》2025年第4期161-173,共13页中国通信(英文版)

摘　　要：Wireless communication systems that incorporate digital twin(DT)alongside artificial intelligence(AI)are expected to transform 6G networks by providing advanced features for predictive modeling and decision making.The key component is the creation of DT channels,which form the basis for upcoming applications.However,the existing work of channel predictive generation only considers time dimension,distribution-oriented or multi-step slidingwindow prediction schemes,which is not accurate and efficient for real-time DT communication systems.Therefore,we propose the wireless channel generative adversarial network(WCGAN)to tackle the issue of generating authentic long-batch channels for DT applications.The generator based on convolutional neural networks(CNN)extracts features from both the time and frequency domains to better capture the correlation.The loss function is designed to ensure that the generated channels consistently match the physical channels over an extended period while sharing the same probability distributions.Meanwhile,the accumulating error from the slicing window has been alleviated.The simulation demonstrates that an accurate and efficient DT channel can be generated by employing our proposed WCGAN in various scenarios.

关 键 词：channel generation channel prediction deep learning digital twin(DT) generative adversarial network(GAN) 

分 类 号：TN929.5[电子电信—通信与信息系统]