An Improved WiFi Positioning Method Based on Fingerprint Clustering and Signal Weighted Euclidean Distance

. 2019 May 18;19(10):2300. doi: 10.3390/s19102300

Algorithm 1: The proposed PL-assisted clustering algorithm.

Input: offline RSS fingerprints

{R S S_{1}, R S S_{2}, \dots, R S S_{N}}

;
position labels of offline RSS fingerprints

L = {L_{1}, L_{2}, \dots, L_{N}}

;
number of prototype vectors Q;
initialized prototype vector set

V = {V_{1}, V_{2}, \dots, V_{Q}}

;
position labels of prototype vectors

T = {T_{1}, T_{2}, \dots T_{Q}}

;
learning rate

α \in (0, 1)

;
1: repeat
2: randomly select an offline RSS fingerprint from the database;
3: calculate the Euclidean distance between the selected

R S S_{i}

and all prototype vectors:
4: find the prototype vector

V_{j}

closest to the selected

R S S_{i}

:
5: if

L_{i} = T_{j}

V_{j}^{'} = V_{j} + α (R S S_{i} - V_{j})

;
7: else
8:

V_{j}^{'} = V_{j} - α (R S S_{i} - V_{j})

;
9: end
10: the prototype vector

V_{j}

is updated as

V_{j}^{'}

;
11: return to line 2;
12: until achieve the maximum number of iterations or the update range of the prototype vectors is very little;
Output: the final prototype vector

{V_{1}, V_{2}, \dots, V_{Q}}

;