Research On Robust Indoor Localization Algorithms Using Low Rank Matrix Completion

Recently,indoor localization has gained unprecedented opportunities with the rapid development of modern communication technology. It is crucial but challenging to obtain a robust and accurate localization approach in the complicated indoor environment which would not keep static due to the NLOS propagation, multipath effects and moving people etc. Robust indoor localization methods based on low rank matrix completion are proposed for the sake of the corresponding backgroud. The main contribute of this thesis is as follows.Firstly, an overview of the tranditional RSS based indoor localization strategies is provided. Combining with the performance evaluation of indoor localization technology standards, the existing problems of RSS based indoor localization algorithms are analyzed. Besides, the analysis process help provide the theoretical foundation and direction for the algorithms proposed in the following chapters.Secondly, the LRMC theory which is a prevalent signal processing method is presented. The feasibility of combing the powerful technique with the indoor localization is also analyzed.Then, indoor localization approaches based on RSS fingerprinting and the structure of the fingerprinting are discussed. The inefficiency and inaccuracy of the conventional methods based on RSS fingerprinting are mainly caused by the time consuming procedure of fingerprinting construction, outliers in RSS measurements and signal receivers’ heterogeneity. In order to improve the disadvantages, a localization approach using multiple-channel SSD fingerprinting based on LRMC and the improved version are jointly proposed. Furthermore, MATLAB simulation results and real data validation based on ZigBee network both demonstrate the feasibility and superiority of the proposed algorithms.Lastly, indoor localization methods utlizing the low rank EDM completion are proposed. Data incompleteness, data inaccuracy and poor environmental adaptability contribute to the poor performance of the conventional localization approaches using the EDM. Considering the low rank property of the EDM, a novel algorithm based on LRMC is proposed to obtain the complete and robust information of the EDM. With full EDM, SDP relaxation based localization approach is proposed to fix the locations of the targets. Besides, simulation results and real data verification show that the proposed algorithms can effectively improve the traditional EDM-based localization methods.