Evaluation of Improved K-Nearest Neighbors for Indoor Positioning System in Real Complex Buildings

Developing indoor positioning systems (IPS) using Received Signal Strength (RSS) received from Wi-Fi sensors based on the Fingerprinting technique has become an interesting topic to many researchers in the positioning and navigation field. These systems, however, are barely tested in real environments such as complex buildings and in real situations where the change of RSS over time occurs frequently. Moreover, the cases tested in an environment close to a real scenario could not provide proper positioning accuracy. In this paper, some methods are proposed to improve the IPS with the KNN algorithm which changes the procedure of selecting the chosen train points. These methods are, also, implemented and tested in a real building with different complexities and an area of 324 m
²
. The results show an improvement of 14% in average positioning accuracy from 2.11 m to 1.82 m.