The Global Positioning System (GPS) allows a device that incorporates this technology to obtain its position in global coordinates. However, GPS does not works correctly inside buildings. That is why some other technologies of Indoor Positioning are being developed. Most of these solutions, both academic and commercial, rely on some kind of radio technology to localize specialized hardware. Modern smartphones contain all sensors needed to implement successfully classic indoor location techniques, such as WiFi or BLE. These mobile phones are also endowed with other sensors like the magnetometer, accelerometer or gyroscope, which can help to improve the robustness and precision of positioning systems. The aim of this thesis is to develop an advanced positioning technology able to adapt to different users and how they carry their smartphones and which does not require external infrastructure different from that which is already common in most buildings. With this aim we had to develop a strategy able to detect the movement of the user using the data obtained from the inertial sensors. This method is robust to differences in hardware and users. Our strategy also takes advantage of the already existing infrastructure of the building (WiFi, BLE), together with the information provided by the inertial sensors of the smartphone (accelerometer, gyroscope, magnetometer) in order to determine the position and displacement of the user in two main scenarios: Indoor positioning for pedestrians and for vehicles.
Keywords: indoor localization, machine learning