Probabilistic RSS Fingerprinting for Localization in Smart Platforms

Abstract

Location-Based Services (LBS) are the major services offered by IoT platforms. Localization techniques are the key to these services. This is possible because of modern electronic sensors with radio interface are embedded ubiquitously in IoT platform. Localization techniques analyzes the radio signal fluctuation of these sensors to build a pattern called signal fingerprint. The localization methods can be device based active localization or device-free passive localization. Passive localization method need not place sensor to the target device in contrast to the active method where each target must have a sensor for localization. It estimates the changes in the radio signal pattern to find the target location.These methods are applicable in traffic surveillance, pervasive computing and smart environment. This approach represents the signal parameters to generate the map. Further the new signal from an unknown object is matched with the map to find the location. The device free localization system performance depends on the radio signal model. This paper proposes a probabilistic radio signal model for map generation and localization. Moreover, this paper compares this probabilistic approach with existing KNN approach using UJIIndoorLoc dataset.