Algorithm 1

Input:  (1) The coordinates of N special APs:   $\left\{(a{x}_i,a{y}_i)|i=0\to (N-1)\right\}$ (2) The boundary of localization space S; Output:  Location sequence of all divided regions constitutes location sequence table; 1:Use CSI measurement to get effective CSI value between ordinary AP and all special APs; 2:for each $i\in [0,n-1]$ do 3:    Collect CSI data between ordinary AP and all special APs; 4:    Denoise CSI data by using Butterworth filtering to filter out the high frequency noise; 5:    Filter out time delay based on the time-domain LOS path identification method; 6:end for 7:Estimate the distance and calculate the location sequence of ordinary AP; 8:Calculate the Kendall coefficient ${\tau }^{\prime }$ between the ordinary AP location sequence and the sequence in the location sequence table by using Equation (6); 9:for each $i\in [0,n-1]$ do   ${\tau }_{max}^{\prime }$ = ${\tau }_0^{\prime }$;   ${\tau }_{max}^{\prime }$ = $max({\tau }_{max}^{\prime },{\tau }_i^{\prime })$; 10:end for   return ${\tau }_{max}^{\prime }$; 11:Choose the centroid represented by the location sequence that has the ${\tau }_{max}^{\prime }$ with the ordinary AP as its location estimate.