Research On Multi-source Fusion Positioning Theory And Method

Location based service(LBS)has attracted more and more attentions,which is considered as one of the three sunrise industries in information technology of 21 th century as well as mobile communication and Internet.Global Navigation Satellite System(GNSS),one of the most competitive positioning techniques,provides reliable positioning and navigation services for global users day and night.However,GNSS cannot work caused by the insufficient coverage of navigation satellite in some positioning applying environments such as indoor environment,area between high buildings in urban,mountain and valley environment,underwater environment,etc.Multi-source fusion positioning technique is an effective supplementary scheme of GNSS.Based on data fusion strategy,multi-source fusion positioning system(MFPS)merges different positioning methods including satellite positioning,wireless communication signal positioning and motion sensors positioning to achieve the best fusion positioning results.Multi-source fusion positioning adopts data integration technology,which increases degree of freedom of positioning data.Furthermore,it can improve the application range of localization and enhance the seamless positioning capability greatly.In MFPS,positioning information is gathered from fusion sources and then transmitted to fusion center(FC)for further processing.The positioning information is merged by using corresponding fusion algorithms to achieve final fusion positioning results in FC.Compared to a single fusion source,MFPS can not only enhance positioning accuracy,but also improve the reliability and robustness of positioning system.Furthermore,the fusion algorithms can be adaptively updated using the feedback and reprocessing of fusion results.This thesis will focus on three aspects including multi-source fusion positioning data preprocessing,multi-source fusion positioning algorithms in different positioning environments,and multi-source fusion positioning effectiveness evaluation.For the positioning information processing of fusion sources in MFPS,a multi-source fusion positioning data preprocessing method is proposed,which aims to design an efficient information processing method of fusion sources.Firstly,a framework of networkside based MFPS is presented,based on which,a data preprocessing algorithm is proposed including data exacting and data gathering algorithm.Data exacting attempts to remove invalid positioning data based on different characteristics of fusion sources.For the data gathering issue,a big data gathering algorithm based on compressive sensing(CS)for fusion sources is proposed,in which the segmental sparse data is transmitted to FC and then the original data is reconstructed from the segmental sparse data in FC.Data preprocessing method can reduce the amount of collected data and improve network transmission efficiency on the premise of ensuring positioning accuracy.For the fusion positioning algorithms issue in MFPS,a factor graph based multisource fusion positioning algorithm(FGMFPA)applied in static positioning situation and an extended kalman filter based multi-source fusion positioning algorithm(EKFMFPA)applied in dynamic positioning situation are proposed in this thesis.In FGMFPA,the positioning information is merged effectively by calculating the soft-information transmitted between variable nodes and function nodes in factor graph.Furthermore,the FGMFPA is applied in Wireless Sensor Network(WSN)localization.The multilateration positioning method and fingerprint positioning method are merged together to obtain final fusion positioning results,which can improve the positioning performance in WSN localization obviously.In EKFMFPA,a smart-terminal based multi-source fusion positioning system model is presented firstly.Pedestrian Dead Reckoning(PDR)and WiFi fingerprint are considered as the fusion sources in this model.Then the theoretical analysis of filter technology adopted in data fusion positioning filed is discussed,based on which,the EKFMFPA is proposed in smart-terminal localization.EKFMFPA merges PDR and WiFi fingerprint together to improve the performance of smart-terminal based MFPS grealty.For the effectiveness evaluation problem in MFPS,a multi-parameter and multi-level model is introduced firstly in this thesis.Five parameters including positioning accuracy,continuity,availability,confidence level,and plug and play capability are adopted to evaluation MFPS in this model.Furthermore the evaluation theory of MFPS is analyzed,based on which,a self-feedback effectiveness evaluation algorithm is proposed.The proposed effectiveness evaluation algorithm contains three phases including initial,fusing and selffeedback phases,and it can not only realize the performance evaluation of MFPS,but also improve the performance of fusion positioning system through feedback information.