Localização em ambientes internos utilizando redes IEEE 802.11

Abstract

This work proposes a method that employs the k-Nearest Neighbors (kNN) machine learning algorithm to determine the location of moving objects indoors. In the test scenario, the mobile object is represented by a Wireless Station (WSTA) that uses Wi-Fi (Wireless Fidelity) technology. In order to estimate the location of the WSTA, measurements were made of the Received Signal Strength Indicator (RSSI), from signals from access points (APs), from specific collection points denoted as points reference points (RPs). In this scenario, in an initial phase of training the algorithm, each RP is used to collect RSSI samples in a process of scanning APs installed in the environment. Also in the training phase, quartiles measurements are used to represent the behavior of these RSSI samples. Subsequently, in the test phase, the training set data, formed by the quartiles, are compared with new data in order to determine the position of the WSTA. In the performance evaluation, it was verified that the proposed algorithm had null error with only four APs e 10 readings per sample with 17.27 seconds of processing time. It is verified that the results with these values are important contributions, which ensures that using the kNN algorithm adopting a dataset summarized with quartiles measurements is a promising method to locate objects indoors.