Transfer Generative Adversarial Networks (T-GAN)-based Terahertz Temporal Channel Modeling and Generating

Terahertz (THz) communications are a promising technology for 6G and beyond wireless systems, offering ultra-broad bandwidth and thus data rates of Terabit-per-second (Tbps). However, accurate channel modeling and characterization are fundamental for the design of THz communications. Relying on channel measurements, traditional statistical channel modeling methods suffer from low accuracy due to the assumed certain distributions and empirical parameters. Moreover, acquiring extensive channel measurement is time-consuming and expensive in the THz band. To address these challenges, a transfer generative adversarial network (T-GAN) based modeling method is proposed in the THz band, which exploits the advantage of GAN in modeling the complex distribution. Moreover, the transfer learning technique is introduced in T-GAN, which transfers the knowledge stored in a pre-trained model based on simulated data, to a new model based on a small amount of measured data. The simulation data is generated by the standard channel model from 3rd generation partnerships project (3GPP), which contains the knowledge that can be transferred to reduce the demand of measurement data and improve the accuracy of T-GAN. Experimental results reveal that the distribution of power delay profiles (PDPs) generated by the proposed T-GAN method shows good agreement with measurement. Moreover, T-GAN achieves good performance in channel modeling, with 9 dB improved root-mean-square error (RMSE) and higher Structure Similarity Index Measure (SSIM), compared with traditional 3GPP method.