RSSI-KNN: A RSSI Indoor Localization Approach with KNN
Xiaoshan Zheng, Ruojin Cheng, Yiming Wang
Abstract
Nowadays, with the rapid growth of wireless communication technique, Received Signal Strength Indicator (RSSI) is not only satisfied in measuring distance, but also be used in indoor location frequently. The article focuses on one technology which is called Trilateration, which measure the energy of the RF signal to determine the distance of the transmitter and the desired point. Furthermore, this article also provides a more precise memoryless method-K-nearest neighbor (KNN), which makes an excellent matching of the test point in the test set through the fingerprinting-localization model constructed for the dataset. Based on a complex indoor scenario with several corners and shelters, this article has made a comprehensive comparison of these two technologies through giving an accuracy and precision analysis in three familiar wireless technologies: ZigBee, Bluetooth Low Energy (BLE) and WiFi. However, according to the experimental results, KNN is significantly better than Trilateration at Indoor Localization. The average of MSE using KNN in three technology was 1.1613m with a variance of 0.1633m. The average of MSE using Trilateration was 2.2687m with a variance of 4.8903m. By comparing the minimum mean square error (MSE) and variance (σ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) when providing different values of k to this scenario, this article gets the optimal is 3 to make the k-value which was chosen won’t lead overfitting or underfitting. Under these circumstances, the MSE and the variance of the indoor scenario can be reduced to 0.9577m and 0.1288m respectively. In terms of the various wireless technology, WiFi has a higher accuracy under Trilateration and KNN, which the MSE and the variance are 1.7109m and 1.8182m.