Toward Robust And Scalable Indoor Trajectory Tracking System: Research And Implementation

As a key enabling technology of location-based services,indoor localization has attracted considerable attention from both industry and academia.Currently,as a mainstream technology in this field,the Wi Fi fingerprint based indoor localization can be deployed without additional hardware.However,the Wi Fi signal may suffer from serious signal attenuation and reflection in indoor environments,which could decrease localization accuracy and impair the robustness of localization system.Besides,the tremendous site-survey overhead of Wi Fi fingerprint based solutions seriously hinders the system's scalability in real world scenarios.On the other hand,although the dead reckoning technique based on the inertial sensors embedded in smartphone can estimate user's displacement,this technique suffers from cumulative error and may gradually lose the tracking ability.With above motivations,we conduct an extensive study on fusion algorithms based on the Wi Fi fingerprint and the inertial navigation.We propose a particle filter based algorithm,which is named ZSSO(Zero Site-Survey Overhead),to improve the scalability of the tracking system.ZSSO can automatically collect Wi Fi fingerprints and perform a fusion localization to track people with Zero Site-Survey Overhead.On this basis,an Iterative Fingerprint Update(IFU)Strategy is proposed to enhance the system robustness.Finally,we conduct an extensive experimental study to validate the effectiveness of the proposed algorithms.Particularly,the main contents of this thesis are summarized as follows:In ZSSO algorithm,we first design a particle filter based tracking method to effectively estimate user's location based on the user's motion information and the map constraints.On this basis,a zero overhead Wi Fi fingerprint collection method is designed to collect Wi Fi fingerprints without the dedicated site-survey.Then,we propose an improved tracking algorithm by incorporating Wi Fi fingerprint localization results,which can further improve the accuracy and integrity of the estimated trajectory.In IFU strategy,a clustering-based method is first proposed to refine fingerprint database by detecting and deleting outliers.Then,an iterative weight update mechanism is designed for Wi Fi fingerprints,which can evaluate the results of Wi Fi fingerprint localization and update the corresponding fingerprints accordingly.Last,we implement the system prototype and give a comprehensive performance evaluation in real-world environments.Specifically,we first validated the effectiveness of the zero overhead fingerprint collection method in ZSSO to prove the scalability of the proposed methods.Then,by conducting performance evaluation in different indoor scenarios,we demonstrated the proposed methods has a robust tracking performance.