Smart Mobile Device-based Crowdsourcing Framework For Indoor WiFi Fingerprinting And Theoretic Analysis In Localization System Deployment

Due to the ubiquitous existence of radio frequency(RF)signals in indoor environment,research work in RF signal-based indoor positioning embraces more possibilities than ever before.Meanwhile,with the trend that mobile internet are beginning to play a dominant role in internet industry,people has increasing demand on indoor location-based service(LBS).These together forms the factor to draw researchers from diverse background into the study of indoor positioning techniques.Many research works have been conducted to utilize RF fingerprints 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.Besides,the constructed fingerprint database is not time dynamic.Therefore,traditional fingerprinting techniques still can not be applied in real scenarios in large scale even the idea first emerged in the year of 2000.With above mentioned,we address indoor positioning issues from two different perspectives.Firstly,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.Secondly,we address the coarse-grained indoor WiFi localization in shopping malls,a large,complex,but profitable application scenario.The two objectives we pursue of the the WiFi localization system are cost-efficiency,where minimum number of access points(APs)are deployed,and incentives,where each shop is well-motivated to follow our scheme to install the APs.For the first objective,we propose a dynamic programming approach to to compute the cost-efficient WiFi deployment scheme.For the second objective,we propose a valuation sharing scheme in cooperative game theory to balance the interest of each shop.Our results shed light on ho cost efficiency can be achieved for WiFi localization system and on how to motivate agents to contribute to WiFi localization system.