Uhf Rfid Localization with Machine Learning Approaches: Distance Estimation From Rssi Data

Abstract

With the development of the Internet of Things (IoT), sensor technology and mobile internet, indoor localization has gained great importance. Indoor localization is the process of locating objects or people in an indoor environment. Unlike outdoor positioning, technologies such as GPS (Global Positioning System) cannot be used effectively indoors.  The reasons for this include attenuation of GPS signals by concrete structures such as walls, ceilings, etc. and inaccurate position calculations due to multiple signal reflection indoors.Indoor positioning is a solution to needs such as fixture tracking and inventory location control in warehouses and stores. RFID (Radio Frequency Identification) technology, which has been used in industry and academia in the past years, stands out in terms of cost, low power demand and easy applicability.

Indoor location systems designed with RFID technology usually use parameters such as received signal strength indicator (RSSI), time of arrival (ToA) and time difference of arrival (TDoA). Studies with RSSI values can find the estimated distance of the RFID tag to the antenna in response to the received signal strength. In this study, RSSI values of passive UHF (Ultra High Frequency) RFID tags were collected from the experimental environment. Time and frequency domain features are calculated from the RSSI values and these features and tag distance values are used to train machine learning models. The machine learning algorithms are discussed comparatively in terms of mean absolute error (MAE) values obtained from the test results.