Research On Multipath Delay Estimation Based On Meshless Compressed Sensing Algorithm

As an important parameter estimation method in high-precision positioning,multipath time delay estimation is a key technology to support new scenarios such as smart cities,autonomous driving,and motion tracking nowadays.However,in the communication environment with dense multipath interference,the traditional time delay estimation method is limited by the influence of sampling network or sparse grid density,and cannot meet the estimation accuracy requirements.The method based on grid-less compressive sensing improves the accuracy of the parameter estimation problem and provides a new solution route for dense multipath time delay estimation.In this paper,we study the dense multipath time delay estimation problem based on the grid-less compressive sensing algorithm.The main work of this paper is as follows.(1)To address the problem that the traditional grid-like algorithms and off-grid compressed sensing algorithms may not be able to discriminate the time delay values at close intervals,we propose to improve the off-grid sparse iterative method using the grid-less method and conduct simulation experiments.The simulation results show that the grid-less sparse iterative method can effectively avoid the influence of sparse grid and improve the estimation performance compared with the traditional algorithm and the off-grid sparse method.When the signal-to-noise ratio is greater than 0d B,using the Chirp signal,the estimation performance of the grid-less sparse iterative method is improved by more than 11%.The simulation experiment also considers using the OFDM signal model for analysis,and it also proves that the grid-less sparse iterative method has a wide range of applications.(2)For the limitation of the accuracy of the grid-less sparse iterative method by the circular iteration threshold,the grid-less optimization method SPA is proposed to be applied to the time delay estimation problem to avoid the influence of the iteration threshold,and the idea of Wasserstein distance is proposed to be applied to the grid-less sparse optimization time delay estimation algorithm for the problem of high time complexity in the SPA algorithm.The simulation results show that the two grid-less sparse optimization methods used can effectively avoid the influence of iteration threshold compared to the grid-less sparse iterative method.When the signal-to-noise ratio is greater than 0d B,in the same simulation environment,the performance of the two methods is improved by more than 30%and 11%respectively,while the complexity of the latter method is reduced by aboutO(K_r ^2.5)compared with the former method,which effectively utilizes the sparse signal characteristics and improves the performance of the two methods.estimated performance.(3)In order to reduce the computation time of grid-less optimization algorithm,this paper designs a one-dimensional convolutional neural network method based on grid-less sparse optimization processing for multipath delay estimation based on the processing of signal by grid-less sparse optimization,in order to avoid the complex solution process of the calculation of time delay parameters in the grid-less algorithm.The network takes the correlation matrix processed by the grid-less optimization algorithm as input,and outputs the corresponding delay parameters after processing by the one-dimensional convolutional neural network.The simulation results show that,compared with the grid-less sparse optimization method,the convolutional neural network used for multipath delay estimation after training can greatly shorten the estimation time while ensuring the accuracy of delay estimation.CNN is about 1/10 of the off-grid sparse iterative algorithm,and about 1/8000 of the grid-less sparse optimization algorithm(SPA),effectively reduce the performance loss of the system.