Research On Technology Of Indoor Pedestrian Localization And Tracking Based On Smartphone

With the development of wireless communication technology and the rise ofubiquitous computing, Location Based Services (LBS) provided by smartphones areplaying an increasingly crucial role in people’s daily lives, especially in indoorenvironment, which highlights their commercial, academic and social values.However, all values of LBS rely on the accurate indoor localization and trackingmethodologies. Though Global Positioning System (GPS) performs well inlocalization outdoors, it does not fit for locating indoors for its distinct natures.After decades of research, indoor localization and tracking improveddramatically. Mainstream approaches of indoor localization and tracking generally fallinto2categories: Pedestrian Dead Reckoning (PDR) and fingerprinting-basedtechnology as well as model-based technology that base on Received Signal Strength(RSS). The former need real-time calibrations to reduce accumulated errors due to theprecision constraint of low-cost hardware. The latter need considerable manual costsand efforts beforehand either to train an accurate range-model or to build a fingerprintdatabase, and need to communicate with an extra server, which increases overheads ofcomputation, storage and communication.In this thesis, based on previous work, an indoor localization methodology isproposed to correct errors of PDR combining with the pedestrian’s RSS data of indoorWiFi APs. This methodology does not exploit RSS absolute values but utilize thechange feature of RSS vectors while walking.3practical issues that may happenindoors, i.e. adverse impact to the trajectory of hand trembling, room estimationmistake resulting from accumulated errors and “hitting” the wall many times unreasonably, are analyzed and studied. Hence, Turn Verifying Algorithm, RoomDistinguishing Algorithm and Entrance Discovering Algorithm are proposed andevoked in the iterative process of Particle Filter (PF), which reduces locating errorconsiderably.On the basis of such algorithms, combining the improved PF algorithm, PDR andfloor plan of the test site, an indoor localization and tracking system named WaP(WiFi-Assisted Particle filter) is designed and implemented. What WaP regards as theinput is the floor plan with the coarse knowledge of which room APs reside in, ratherthan the exact locations of every AP.The prototype of WaP is implemented in Android smartphone. As alight-weighted locating system, it is not necessary to conduct mass computation in anextra server, and typically400particles make the performance satisfied, which savesoverheads of both communication and computation. Extensive experiments conductedin a1362m2indoor office environment reveal that WaP achieves an average locatingerror of0.71m, a sub-meter level locating error.