Artificial neural network-based indoor localization system using smartphone magnetometer

In recent years, localization systems have become an interesting research topic since they present a key factor for location-based services. Although the global positioning system (GPS) is widely used for outdoor positioning, it does not provide the same accuracy in indoor environments. As a result, many alternative indoor positioning technologies have been investigated to tackle this problem during the last few years. However, most existing approaches (e.g., Camera, WiFi, and infrared-based methods) for indoor localization mainly rely on infrastructure, which is expensive and not scalable. Today, the expansion of smartphones possessing a variety of embedded sensors helped develop a precise indoor localization that can meet the requirements of location-based services. This paper proposes an indoor localization system based on magnetic field sensed via a smartphone magnetometer. Anomalies caused by the presence of ferromagnetic materials are used as signatures and fingerprinting to identify different locations. Accordingly, an Android application was developed to build a geomagnetic fingerprinting database in the corridor of Polytech Nantes, France. 7600 signatures were stored in the database, cleaned, and standardized. 70% of data is used to train and validate different multi-output regression models. Extensive simulations are conducted to find the suitable model and to tune model’s hyper-parameters. Once the model is configured and trained, remaining unseen data is used to evaluate the accuracy of the proposed system. Obtained results demonstrate that ANN is the most accurate model with Mean Absolute Error (MAE) equals 0.13m for the studied environment. Only 8% of testing samples have errors higher than the MAE. Moreover, the proposed indoor localization system can locate the user without prior knowledge of his initial position.