Localização de dispositivos móveis usando roteadores com antenas direcionais e classificação de dados

Abstract

WiFi Location systems are increasingly present in our days because of the great ease that the devices can provide to users, such as mobility, independence of factors that limit and cause discomfort usability as cables, for example, and also be more convenient for the user to use something easy to handle. Apart from these points, it is known that the WiFi devices are increasingly developed, providing more opportunities for use in the routine of users. WiFi Location Systems perform device or people discovery in a given environment based on power signal measurements obtained by exchanging packets between devices that compose such systems. This paper proposes a Location WiFi method using multiple routers with directional antennas in the location environment. Routers with directional antennas are used in this study due to its ability to limit signal coverage for a given area, unlike systems with omnidirectional antennas. The impact that a packet sent by a device with unknown position generates in directional antennas installed at the location system can provide different RSSI values. Through the system offline stage where the training part is done, a database is created with all RSSI values obtained from each packet received by directional routers and positions from where the packages were sent. This database is used as knowledge for the classification algorithm so that it can estimate from where a package was sent, only based on RSSI values obtained at each directional router. Two classification algorithms were compared in this proposal to analyze the performance characteristics of each one in any location enviroment. They are the algorithms kNN and Naive Bayes, both widely used in the literature. During the the training part of the localization system composed by directional routers, it was also analyzed how to organize the routers to get results with greater location accuracy. Thus, it was possible to analyze the organization of the routers in the location environments, being together and pointing in different directions, and separated and focused to the center of the localization environment. ix Another important point of this proposal is the application of the localization system based on two types of location: Precise Location and Contextual Location. The precise location allows to estimate the exact position of a device in the axes x and y of a particular environment where it is installed. And the contextual location allows to estimate the position of a particular device based on the analysis of the context of the environment where it is inserted. All data collected during the experiments were analyzed and tested, and show that the system proposed in this paper can be used as a WiFi positioning system for indoor environments as an alternative to the system commonly found in the location literature, composed of omnidirectional routers. According to the structure of the location where the system was inserted and the organization of directional routers, it was possible to obtain very similar results compared to the location system composed by omnidirectional routers. This can be seen from the results obtained in a small location environment composed by the ICOMP laboratories where the system with directional routers and data classification from the kNN algorithm, has achieved 84.26% of accuracy in positioning compared with 83.93% achieved with directional routers together using the same classification algorithm. Another interesting point seen in the results of this work is that the use of Naive Bayes algorithm tends to be better when the system uses directional routers rather than omni-directional one. This happens because this algorithm can analyze as the direction of each antenna impacts the characteristics of signal reception in a given environment. This can be seen for example in the results obtained on a larger localization environment composed of ICOMP teachers rooms where the positioning system with scattered directional antennas got 80.16% accuracy in positioning estimation compared with the system composed by omnidirectional antennas that got 71.68%. Thus, we shall show that the tracking systems can be composed of directional routers as an alternative to the WiFi location conventional system.