Single-node WiFi Indoor Positioning System Based On RSSI And CSI Fusion

The traditional WiFi indoor positioning technology based on Received Signal Strength(RSS)has the disadvantages of large fluctuation of WiFi signal strength with time,sensitivity to environmental noise,serious multipath attenuation and low information dimension.Therefore,a single AP can only distinguish meter-level positioning.In recent years,Channel State Information(CSI)can be extracted from physical layer,which can represent the transmission process of signals in finer granularity.It opens up a new space for WiFi-based high-precision indoor positioning.According to the previous researches,this paper proposes an indoor positioning technology based on the fusion of RSSI and CSI.The main work and innovations are as follows:1.This paper introduces the technical model of WiFi indoor positioning,and points out the advantages of using WiFi fingerprint scheme.2.The characteristics of RSSI and CSI are compared and analyzed from different time,frequency and environment.It is concluded that CSI is more stable and more obvious in different locations or environments.Therefore,CSI is more suitable for accurate positioning.3.From the principle of WiFi receiver,the error model of CSI is deeply analyzed,and a solution is proposed for the error model.Among them: the innovative Kalman filter is used to smooth the CSI amplitude to reduce noise interference;the linear phase of CSI is corrected by using unwraping,linear transformation and phase difference based probability statistics;Levenberg Marquart algorithm is used to analyze the error and extract the parameters of the nonlinear phase of CSI.The experimental results prove the correctness of the algorithm.4.We implemented and built a real-time positioning system based on the fusion of RSSI and CSI.This paper uses the Atheros 9380 network card and the open source OpenWRT router to obtain the original CSI and RSSI data;In the pre-processing stage,the CSI amplitude and phase are corrected:CSI amplitude is innovatively smoothed by filtering.And we use quadratic clustering based on Time-Reversal Strength quadratic clustering to eliminate malformed data and reduce data redundancy,which effectively reduce the amount of data stored and improve the quality of fingerprints;In the online phase,the KWNN algorithm of RSSI is used for rough positioning to reduce the number of online CSI,and The Combined Time Reversal Resonating Strength(CTRRS)algorithm of the single AP multi-link fusion is used to accurately locate the CSI.In real-time network communication server,one loop per thread model of multi-threaded Reactor is adopted to ensure efficient communication.We also innovatively designed a two-threshold method on the client and server to provide a reliable positioning mechanism.5.This paper analyzes the experiment from several aspects.The indoor positioning system based on the fusion of RSSI and CSI,which uses TR clustering and RSSI for rough positioning,can effectively reduce about 1/3 of fingerprint reference points,and reduce positioning time and maintain positioning performance.After multi-antenna fusion,the effective bandwidth is extended,and finally we use the CTRRS algorithm to achieve the second-level positioning with an accuracy of 90% in the line-of-sight and non-line-ofsight environment.