A study of Machine Learning Techniques for Multi-Floor Indoor Localization using Wifi Fingerprinting

Indoor localization has recently increased interest due to the potential wide range of services by leveraging the Internet of Things (IoT) and ubiquitous connectivity. Different techniques, wireless technologies and mechanisms have been proposed in the literature to provide indoor localization services to improve the services provided to the users. The constraint with the commonly used fingerprinting approach is the high variation of RSSI values, resulting in erroneous location estimation. The machine learning approach is a new addition to the fingerprinting system that aims to solve this problem. This thesis adopts a grid-based approach to train different machine learning models with an end-to-end pipeline to autotune the model's hyperparameter. This study applies three different machine learning approaches of classification, regression and cascaded models. The evaluation of classification models is benchmarked against the publicly available dataset UJIndoorLoc (UJI) with an accuracy of 88.82%. Moreover, this thesis applies the trilateration methodology for optimization of access points location.