Indoor Localization Using Low-Rank Matrix Recovery and Dynamic Parameter Estimation Algorithm With Commodity Wi-Fi

In the field of integrated sensing and communication (ISAC), achieving high-precision location information based on channel state information (CSI) has emerged as a research hotspot. Traditional subspace-based parameter estimation methods suffer from performance degradation when some sampling points are missing due to a heterogeneous interference source. Besides, the majority of methods ignore the influence of outliers on the precision of estimation results when identifying paths. This article proposes a CSI purification algorithm based on the alternating direction method of multipliers (ADMM), which exploits the sparsity of the CSI matrix in conjunction with low-rank matrix recovery (LRMR) techniques to effectively mitigate interference and enhance path resolution. A parameter estimation method for dynamic outlier removal is adopted, which combines the matrix pencil (MP) and density-based spatial clustering of applications with noise (DBSCAN) algorithm to dynamically adjust the input parameters to achieve the removal of outliers. We performed indoor measurements and showed that the CSI purification algorithm was able to reduce the median angle of arrival (AoA) error from 8.39° to 6.24° when half of the sampling points were missing. Besides, the addition of the dynamic outlier removal algorithm improves AoA estimation performance by 4.1%. In terms of localization measurement, the median localization error of the proposed system in the activity room and the meeting room is 0.39 and 0.41 m, respectively. In the case of missing half of the data, the positioning accuracy of the proposed algorithm is 0.47 m, which is better than that of the mainstream algorithms.