Graphslam-Based Crowdsourcing Framework For Indoor Wi-Fi Fingerprinting

Min Zhang,Ling Pei,Xiaotie Deng

DOI: https://doi.org/10.1109/UPINLBS.2016.7809951

2016-01-01

Abstract:Due to the ubiquitous deployment of WiFi Access Points(APs) in indoor environment, many research works have been conducted to utilize WiFi Received Signal Strength(RSS) for the purpose of indoor localization. Most of such methods require large amount of training data to guarantee satisfying accuracy. However, the process of site survey is really time-consuming and labor intensive. In this work, we propose to combine inertial sensor -based user motion measurement and sensed WiFi signals to enable WiFi fingerprint database construction in a Crowdsource manner. Pedestrians whose smart phone are equipped with our software could contribute his walking trace and WiFi signals to our system without explicit effort. User trajectories are modeled with GraphSLAM which requires a certain number of landmarks for better calibration. We investigate the characteristics of signal similarity in indoor space and on this basis propose to utilize virtual WiFi landmarks for GraphSLAM optimization and we design a two-step algorithm to map these landmarks automatically to fit in with Crowdsourcing settings. Finally, we choose the kNN algorithm as the localization algorithm. Our proposed method frees surveyors from the hard work of site survey in whole environment and at the same time yields a competitive localization accuracy.