Research On Time Reversal Based High Precision Indoor Localization Technology

In recent years,as the number of devices connected to internet gradually increases,kinds of internet of things based applications bring much convenience to people's life.Wherein,location information is one of the most important prerequisites.Hence,how to efficiently obtain precise location information has become the key issue to prompt the development of localization industry.At present,in outdoor localization,there are several mature solutions such as GPS,Beidou,etc.However,complex indoor layout and structure,rich multi-path,scattering and shadow effect make indoor localization face a big challenge.Traditional indoor localization methods intend to remove the impact of multi-path,and extract useful information from the Line of Sight signal,which is inevitably influenced by the rich multi-path in indoor environment.Inversely,time reversal based indoor localization technology could fully utilize the location-specific characteristic of rich multi-path to achieve high resolution localization.Based on analysis of related influence factors of time reversal indoor localization,this dissertation is aimed at bandwidth limited scenario,relatively dynamical scenario and passive localization scenario,and systematically studies high resolution indoor localization.The research's emphases and innovations include:1.Based on previously created the experimental platform,we study the impacts of environment,bandwidth and number of antennas on the performance of time reversal based indoor localization.To further improve the performance of time reversal indoor localization with limited bandwidth,by introducing location-specific characteristic of channel state information(CSI)phase,we present a high resolution time reversal based indoor localization method.In the offline stage,to minimize the storage space of fingerprint database,we apply a density-based spatial clustering algorithm to adaptively obtain fingerprints for each location.In the online stage,by fully utilizing the location-specific characteristic of CSI amplitude and phase,we raise Time Reversal Resonating Strength(TRRS)calculation method which synthesizes both the amplitude and the phase to achieve more accurate localization without requiring more storage space for fingerprints than the original time reversal method.At last,practical experiments verify that the accuracy of the proposal is as higher as 97.5%,which is obviously higher than other baselines.2.Considering that the validity of fingerprint database would be declined with environmental changes,by introducing deep learning into time reversal indoor localization,we put forward a deep learning based self-calibration time reversal indoor localization method.In the offline stage,taking into account that CSI is consist of amplitude information and phase information,based on deep learning theory,we respectively adopt two auto-encoders to train the amplitude information and phase information without labels,and after training,we could obtain the auto-encoders which store the features related to environment.In the online stage,to mitigate the impacts of environmental changes on real-time measurements,we use the trained auto-encoders to adaptively calibrate the amplitude and phase of real-time measurements,respectively,and achieve highly robust localization based on the modified TRRS.Practically experimental results show that the proposal could remove 36.51%,25.0%and 20.49% localization errors cause by environmental changes in the environments with window opened-door closed,window closed-door opened and moving people.3.Considering the scenario of practical application in which it is not convenience for user to carry device,we come up with a highly accurate weighted time reversal passive indoor localization method.By introducing time reversal technology into passive localization,we could achieve high resolution localization results.At the same time,by fully taking advantage of multiple antennas' information,we propose a confidence level based weight calculation method to improve the accuracy of indoor localization.Experimental results show that using2/3 wireless communication links for localization could improve 11.11%/14.81% localization accuracy when comparing with using 1 wireless communication link for localization.To improve the robustness of passive indoor localization system,we present an anti environmental disturbance measurement self-calibration scheme.Finally,we validate the effectiveness of the proposal by conducting practical experiments in dynamic environments.