STUDIO E VALUTAZIONE SPERIMENTALE DI TECNICHE DI RADIO FINGERPRINTING PER NAVIGAZIONE INDOOR CON DISPOSITIVI MOBILI ANDROID. IL PROBLEMA DELLA RACCOLTA DATI RISOLTO MEDIANTE TECNICHE DI GENERAZIONE VIRTUALE DI FINGERPRINTS

Precise localization of smartphones in indoor environments is a challenge. Accurate outdoor localization has been exhaustively implemented through Global Navigation Satellite Systems. On the contrary, indoor localization still continues to be a challenge even though it has been exploited during the last several years. Many indoor localization techniques have been explored in the scientific literature. Most of them require the installation of dedicated infrastructures that is expensive to perform and hard to scale. Indoor localization systems that reuse the building existing infrastructures, on the other hand, can provide high localization performance but require a large effort for the fingerprint database construction. This task is time-consuming and labor-intensive [1]. The excessive time to perform fingerprint data collection affects the popularity of fingerprint-based localization. In my internship, I developed three methods for fingerprints augmentation, namely WF- LERP, WF-DCGAN, and WF-RND, to truly expand the fingerprinting database while reducing the human effort, by virtually increasing the number of fingerprints collected at each location. The proposed solutions reduce the time expense involved in the data collection phase of a factor 20, or, under the same time, increase the accuracy performance of the indoor positioning algorithm up to the 52%. Tests are conducted with fresh fingerprints and in presence of scans that suffer a 4-months time-aging. Extensive experiments and results prove the effectiveness and the advantages of the proposed methods over the existing state-of-the-art solutions.