Prediction Of Key Parameters Of Wireless Channel For UAV Platform Communication

In recent years,the development of unmanned aerial vehicle(UAV)technologies has been very rapid,which has profoundly affected all aspects of social development.The original single UAV’s mode of performing simple tasks could not fully meet the needs of UAVs to perform complex tasks,and UAVs perform collaborative tasks become an inevitable development trend.When UAVs cooperatively perform target detection and recognition tasks,they often need to share high-definition and even virtual reality-level image data,which has led to the need for large-bandwidth mobile communications among UAV platforms.With the deepening of research millimeter wave(mmWave)has shown a great potential in large-bandwidth mobile communications,which can meet the large-bandwidth requirements of UAV platform communications.In UAV mm Wave communications,the highly dynamic nature of UAV platforms causes rapid changes in wireless channels.At this time,the traditional pilot-assisted channel estimation method to obtain channel state information(CSI)between UAVs requires high-density insertion of a large number of pilots to ensure the accuracy of channel estimation,however,the rapid change of UAV channels will make the channel estimation value overdue.To solve this problem,this thesis proposes a novel adaptive-structure extreme learning machine(ASELM).The algorithm can reduce the error of channel prediction and support UAV mm Wave communication.In addition,a sliding window prediction mechanism(SWPM)is proposed to achieve online channel prediction.The detailed research work is listed as follows:1)This thesis gives the background and significance of UAV mm Wave communication,summarizes the current research status of channel prediction,comprehensively analyzes the research status of autoregressive(AR)models,and echo state networks(ESN)basic algorithm principles,and simulation analysis of the application of these algorithms in channel prediction.2)For the UAV platform communication scenario,this thesis designs the ASELM channel prediction algorithm.This algorithm adaptively adjusts the number of ELM hidden neurons to adapt to channel changes.Simulation results show that the proposed algorithm achieves smaller prediction error than AR algorithm,ESN algorithm and traditional ELM.3)In order to achieve online channel prediction,this thesis further proposes the SWPM mechanism to further enhance the ASELM algorithm.The ASELM is trained by effectively multiplexing the predicted CSI values to predict subsequent CSI.After the error is reduced,the prediction value will continue to be used for the next channel prediction,thereby realizing online continuous channel prediction.Simulation results show that the algorithm can track channel changes online and perform long-term and continuous channel prediction.