Implementation Of Link-Level Simulation Platform And Research On Channel Estimation For UAV Communication

As an important carrier for transmitting and exchanging information such as images,audio,and control signals,unmanned aerial vehicle(UAV)communication systems play a crucial role.Considering that link-level simulation is a powerful tool for modeling and evaluating UAV communication systems,and channel estimation is a key step in the physical layer design of UAV communication systems,this thesis focuses on the implementation of the UAV link-level simulation platform and the research of channel estimation.Firstly,a physical layer link-level simulation platform for UAV scenarios based on the open-source 5G K-SimLink and the C++mathematical operation library Armadillo is constructed.On the one hand,some necessary modules are implemented and modified to achieve the basic functions of the link,such as channel simulation,multi-stream transmission,channel coding,frame structure,orthogonal amplitude modulation,etc.On the other hand,channel estimation and precoding modules are improved to enhance the performance of the link.This simulation platform can simulate end-to-end transmission links with single-input single-output(SISO)under the additive Gaussian white noise channel and multiple-input multiple-output(MIMO)under the clustered delay line(CDL)channel.By simulating and analyzing the performance of the links with different modulation coding schemes,theoretical guidance and technical support can be provided for the design of UAV communication systems.Secondly,a multi-slot conditional generative adversarial network(cGAN)based on Swin Transformer is proposed for unquantized channel estimation in the SISO scenario.Specifically,the proposed scheme can learn temporal correlation features from continuous channel slots through three-dimensional convolution,and extract deep channel features through Swin Transformer to improve the accuracy of channel state information.In addition,a data augmentation method,which randomly extracts the narrow-bandwidth sub-channel response from the full-bandwidth channel response,is used to decrease the computational cost for offline training.The results of the simulation show that the proposed scheme performs better than the LMMSE method and other deep learning methods.Additionally,to extend the proposed scheme to the MIMO scenario,two extension methods are proposed and validated by simulation.