An Environment­ Adaptive Approach for Indoor Localization Using the Tsetlin Machine

Abstract

Indoor positioning is a challenging task due to the small scale of area and the complex electromagnetic environment. Among different distance measurement schemes, Received Signal Strength Indication (RSSI) readings are commonly used in proximity and localization applications such as in BLE and Wi-Fi, because of the low power consumption and simplicity of retrieving this information. There are several approaches for RSSI based indoor localization, among which the deep-learning based models trained with fin-gerprinting data can achieve far superior localization accuracy compared with orthodox approaches, such as trilateration. However, fingerprinting requires extensive manual labor during the offline data collecting phase for training and cannot adapt well to changes in the environment.