Radio Map Construction From Crowdsourced WiFi Samples For Indoor Localization

With the development of Internet and the popularity of smartphones,indoor location information plays a very important role in numerous industrial and commercial applications.In the last decade,indoor localization has been extensively studied,and various localization schemes based on different technologies have been proposed.Among them,WiFi fingerprinting technique is a research hotspot.In this thesis,in order to reduce or eliminate the burden of traditional radio map construction,a radio map construction scheme with light site survey sample is first proposed.A three-layer localization method based on virtual fingerprints generated respectively by polynomial fitting,virtual source fitting and Gaussian process regression is adopted.Then,a radio map construction scheme based on crowdsourced samples for indoor localization is proposed to solve the inaccurate sample annotation,unequal sample dimensionality,nonuniform spatial distribution and measurement device diversity.In response to the above problems,our solution includes abnormal sample detection,AP selection,fingerprint interpolation and device calibration.Finally,for the two schemes mentioned above,we collect WiFi data in different scenarios and verify those schemes with the measured data.The experimental results show that the virtual fingerprint not only reduce the number of training fingerprints but also improve the accuracy of localization in the first schemes,and the proposed radio map construction scheme is effective and has improved the localization accuracy,and the average localization error can achieve around 1.5m.