UAV Positioning Based On Segment

In response to the increasing number of UAV"hack fly"incidents,regulating UAVs has attracted attention everywhere.UAVs positioning has become an important part of regulating UAVs.Due to the small size of the UAV and the fast flying speed,the traditional positioning method is no longer applicable.It is necessary to research the positioning of the UAV based on the characteristics of the UAVs.This paper constructs UAV positioning from both theoretical and experimental directions.The information on UAV propagation channels through antennas was collected,and characteristic parameters of UAV communication signals in the propagation channels were extracted.The effect of signal parameters on UAV positioning was analyzed through experiments.A segmented path loss and shadow fading model was constructed in this article.UAV signals was clustered by different characteristics.Finally,the radio map was reconstructed by the collected UAV signals to achieve UAV positioning.The main research results of this article are as follows:?1?The UAV signal transmission and reception platform was built.PCI and RSRP were extracted from the UAV propagation channel as experimental parameters.PCI is determined by the placed antenna,and the value of RSRP becomes smaller and smaller as the distance increases,so that it exceeds the receiving range of the antenna.?2?The relationship between the path loss,shadow fading,multipath and distance are studied.The path loss decreases linearly,and the slope of the drop is ten times.Based on this model,a segmented path loss and shadow fading model was proposed.The propagation information was divided into more segments.The variance of?_dB was reduced,and the accuracy of the UAV's channel model positing was improved.Iterative algorithm and Lagrangian method were used to convert non-convex problems into convex problems to solve,thereby solving the MLE problem of parametric?.?3?Through empirical conclusions,it was found that the data set of the received signal of the UAV is split into a training set,a verification set and a test set.The splitting rule is to split according to 7:3:1 or 6:2:2.Different splitting methods are adopted for the data volume with different sizes,and simple methods of leave-out verification,K fold verification and repeated K fold verification are adopted.Through simulation experiments,reducing the model size,adding L1 regularization,L2 regularization,and dropout regularization penalty function functions can all improve the model overfitting,especially L2 regularization is the best.?4?Observe and analyze the data in the collected propagation channel,and get an empirical formula.The radio map was reconstructed by the DC segmentation method,and KNN and SVR methods were used as a reference to reconstruct PCI and RSRP for the same data.Experimental results show that compared with KNN and SVR,DC segmentation method has smaller RMSE reconstruction error and higher accuracy in 10000 training samples,thus verifying the advantages of DC segmentation method in UAV positioning.?1?The RMSE values of PCI and RSRP in the SVR method were 0.6011 and 2.8581,respectively.The DC segmentation method reduced the RMSE of PCI by 54%to 0.2760,and the RSPE of RSRP decreased by 35%to 1.8634.?2?The RMSE values of PCI and RSRP of the KNN method are 0.3301 and 2.1325respectively.Through the DC segmentation method,the RMSE of PCI is reduced by 16%to0.2760,and the RMSE of RSRP is reduced by 13%to 1.8634.