Smart Probabilistic Fingerprinting for Indoor Localization over Fog Computing Platforms

Indoor navigation and localization are becoming fundamental services nowadays. WiFi-based solutions such as FingerPrinting (FP) are the most widely adopted techniques for positioning and provide better results if compared to other approaches. They require to compare WiFi Received Signal Strength (RSS) with an pre-computed radio map called fingerprint. The recently proposed Smart Probabilistic FingerPrinting (P-FP) algorithm reduces the computational complexity of the traditional FP approach without any accuracy detriment. On the other hand, fog computing has emerged as a new promising paradigm in the recent years, which extends traditional mobile cloud computing capabilities towards the edge of the network and enables location-aware services. In this paper we propose to offload Smart P-FP computation over a fog platform exploiting a novel distributed algorithm. Performance evaluation validates the effectiveness of the proposed approach with the analysis of: i) the amount of power saved and ii) the efficiency of candidate selection process for offloading. Having 2 or more devices in the vicinity contributing to the computation makes offloading beneficial from a power standpoint. The offloading effectiveness increases with the number of devices willing to contribute and the amount of data to be transferred. Power savings can be as high as 80% if compared with local computation.