Research On WiFi Positioning Algorithm Based On Location Fingerprint

WiFi (Wireless Fidelity) technology is a kind of wireless communication technology, withcharacteristics of high transmission rate, easy networking, strong mobility. With arrival of theera of Internet of things, more and more attention is paid to the location based service(LBS),which bursts out great vitality in many fields, such as emergency rescue, medical care,personalized information delivery and so on. However, the traditional positioning technology,such as GPS (Global Positioning System) positioning system, is not suitable for the complexchanges of indoor environment. With development of wireless networks, WiFi technologybecomes the preferred technology for wireless LAN network. Indoor positioning technologybased on WiFi signals has advantages of wide range of use, low cost, portability and becomes aresearch hotspot in field of indoor position sensing.Indoor localization usually takes advantage of algorithm based on RSSI. Positioningalgorithm based on RSSI can be divided into range-based and range-free localizationalgorithms. Range-based algorithm has great dependence on indoor propagation loss model,which has to be calculated in real environment each time. Location fingerprint positioningalgorithm is a kind of range-free algorithms and just needs measurement of RSSI to buildfingerprint database. Every fingerprint has only one location information. Unknown fingerprintmeets with fingerprint database (FPDB) to estimate location. Location fingerprint positioningalgorithm realizes easily.In general, location fingerprint algorithm based on WiFi has two stages. The one stage isoffline sampling.The other is to calculate position. Studying at the the positioning principleof fingerprint localization algorithm, this thesis points out factors which may produce error inlocalization process. Besides, the thesis makes a deep and systematic study on error factors andpoints out the limitations of the existing location estimation algorithm. On the basis ofanalyzing and summarizing the existing methods that reduce the influence of errors on locationfingerprint algorithm, this paper proposes a improved location fingerprint algorithm based onk-means clustering and WKNNSS. Proposed algorithm initializes fingerprint database withrepeated measurement of average value through the acquisition of RSSI, and then analyzes the trained fingerprint database using k-means clustering, removing some bad fingerprint. In orderto decrease FPDB’s influence on location, unknown fingerprint matches training fingerprintdatabase, which reduces searching space, improves accuracy of FPDB. The positionestimation stage, the existing nearest neighbor algorithm has been improved. This paperpresents a weighted k nearest neighbor algorithm and introduces a new calculation methodof weight coefficient.Finally, this paper has researched on the influence of measured AP number to locationestimation. Through algorithm comparison and experimental analysis, the validity andpracticability of the proposed solutions are demonstrated. By comparison with originalalgorithm, the advanced algorithm can reduce location errors effectively and has more stableperformance without adding any cost.