Modeling And Algorithm Implementing For RSSI Fingerprint-based Wi-Fi Indoor Localization

The rapid development of Internet of Things and rising of Location Based Service have caused a new upsurge of researches on wireless localization.Combining the advantages of both satellites and mobile communication networks,current outdoor positioning systems have achieved an accurate positioning with very little time taken.Yet people's calling for a much more higher accuracy when it comes to indoor localization to meet commensurable requirements of certain applications,coupled with the fact that signals transmitting in limited space are very susceptible to walls,ceilings,and obstacles,induces difficulties to suffice.In related research,RSSI Fingerprint-Based Wi-Fi localization exploits raw data processing methods,offline database training approaches,online position estimation and real-time tracking algorithms,both of which involve a wealth of mathematical theory and have considerable room for improvement,acts as a hotspot that receives extensive attention from colleges,research institutes,and enterprises.In RSSI Fingerprint-Based Wi-Fi localization system,you can flexibly set reference points on the fingerprint plane,and there is no need of a resort to geometric methods for solving the coordinates,thus errors introduced by initial-biased value and iterative calculation are avoided.In practice,however,RSSI values read out in the very same position by apparatus provided by different suppliers and chipset manufactures,which use a variety of measurement standards,are different in a large scale,besides,these values themselves fluctuate from time to time with temperature,co-channel interference,and some other factors.Thus,the thesis carries out an RSSI collection test of one single AP and analyzes the stability of the values.Data visualization shows the propagation characteristics of Wi-Fi signal power in practical environment.Due to limited experiment conditions,theoretical researchers often cannot effectively get rich and representative RSSI values that are reliable enough to verify the performance of proposed positioning algorithms.To address this issue,the thesis designs and implements an RSSI fingerprint space generating model based on 2-ray tracing methods.The model utilizes a concept of electromagnetic radiation,which derives from Friis equation and Maxwell equations to calculate RSSI on direct path of transmission in terms of ray tracing,and it can also achieve the RSSI on reflecting path by introducing a reflection coefficient,finally the model uses one direct ray together with six reflect rays,of which there are two different types and a total number of seven individual ones,to establish RSSI fingerprint database for a parameter-known indoor scene.This model can flexibly provides a vast number of RSSI that are highly reliable for algorithm verifications,at the same time the model trail and error experiments of an exactly the same scene,which is conducive to the continuation and deepening of theoretical research.The drawback of the poor anti-interference ability of online localization algorithm based on K Nearest Neighbors,which moreover procures failings to leverage a prior information as a result of independent estimating of each position's coordinates,sets a limit to its precision.To solve this problem,the thesis first analyzes the model of hypothetical movement at a constant speed,after that,Kalman filtering is introduced into KNN-based online positioning,and implements a real-time tracking algorithm of which the error converges with the growing of traces.Finally,the thesis verifies the effectiveness of the proposed RSSI generating model and the joint positioning algorithm through intensive experiments,meanwhile,works that have been accomplished in the thesis are concluded in a retrospective way and a prediction for possible tech trends of future indoor localization is made.