Algorithm 1. BSA information detection and coverage identification

Input: $D=\left\{d_i\right|1\le i\le S\}$

1. for eNBID ∈ {eNBIDs in D} do

2.    Remove abnormal samples according to Equations (1) (2);

3.    Calculate the geometric center $(\widehat{x},\widehat{y})$ according to Equation (3);

4.    Determine tob according to Equation (4);

5.    if (tob==sectorized) then

6.         for cellID ∈ {cellIDs of the eNBID} do

7.         calculate initial BS location (x0, y0);

8.         estimate the azimuth θ θaccording to Section 3.2;

9.         $(\tilde{x},\tilde{y})\leftarrow$ geometric center of (x0, y0) for all the co-site cells;

10.   else $(\tilde{x},\tilde{y})\leftarrow (\widehat{x},\widehat{y})$;

11.   Estimate cell range $\mathsf{\gamma }$γaccording to Equation (5);

12. Divide the coverage area into grids {Gi, i = 1~N};

13. for each grid whose number of samples > Ts do

14.   if (average signal strength of the grid<−110dBm) then

15.     mark the grid as “weak coverage grid”;

16.   else

17.     for each cell who has samples in the grid do

18.        calculate Si and Ri, and Di;

19.        calculate ${\tilde{S}}_i$ Siaccording to Equation (6);

20.        calculate cell coverage probability Pi for each cell;

21.        if (Pi > Tc) then

22.           mark the cell i as “dominant cell” of the grid;

23.        else mark the grid as “non-dominant grid”;

24. Smoothing the judgment of dominant cell for each grid;

25. Update {eNBID, cellID, tob, $\tilde{x}$, $\tilde{y}$, $\gamma$ , $\theta$} to BSA;

26. Output coverage identification results of each grid.