| Location based service(LBS)plays an important role in many fields,such as public security,terrorism combat and stability maintenance,emergency rescue and so on.The demands for high-precision localization are boomed especially with the development of pervasive computing,the fifth generation of mobile communication and other key technologies.Traditional positioning methods based on single source would have poor performance in complex environments due to lack of information,resulting in unable to provide high accuracy and availability positioning.Multi-source information fusion-based positioning(MSIFP)technologies could integrate satellite navigation,wireless network localization and sensor-based positioning effectively,so as to achieve more accurate and robust location estimations.Therefore,MSIFP has become one of the most important technologies in the field of LBS applications.The popularity of the intelligent mobile devices has greatly broadened the application scenarios of LBS.However,the variety of the type and accuracy of the built-in sensors brings many challenges to high accuracy and availability positioning.The MSIFP for the application of intelligent devices still faces several key problems,such as the high representation for the spatial resolution of positioning features,the accurate analysis and compensation of the errors caused by body-shadowing,the high precision fusion for multi positioning features,etc,which would lead to poor positioning accuracy and robustness.Aiming at these problems,this paper deeply studies the key technologies for MSIFP towards intelligent mobile devices,such as the association representation of multi-dimensional positioning features,the multi-feature collaboration-based compensation of the shadowing error,and the multi-level accurate fusion of the positioning information.Meanwhile,theoretical researches and experiments have been carried out.The main contents and innovations of this paper are as follows:1.In view of the high representation problem of the spatial resolution for fingerprinting features,the mechanism of the structure for fingerprint and the calculation for feature similarities to the spatial resolution are firstly analyzed.A feature representation model based on the association of multi-frequency and the self-consistent of space-frequency-phase is built,and a fingerprint feature named CCF,is then established under the fact that the spatial sensitivities of the amplitude and phase features in channel state information(CSI)are complementary.The fingerprinting method based on CCF and its dynamic similarity calculation is proposed to achieve high positioning accuracy.In order to reduce the computational complexity of the CCF matching,an optimization method for the selection of reference points is established based on the correlation of local signal strength,which could restrict the number of reference points involved in the matching stage.The experimental results show that the spatial resolution of the proposed CCF is improved by about 5.3%and 25%compared with the amplitude and phase features,respectively.Meanwhile,the positioning accuracy is also improved based on the proposed method when comparing with the method based on Euclidean distance and time reversal resonating strength.2.To address the accurate analysis and compensation problem of the errors caused by body shadowing in signal strength measurements,the relationship among the shadowing error,the distances between device and anchors,and the positioning accuracy is derived theoretically and analyzed through simulations.An adaptive distinguish strategy for body shadowing state is then established based on the heading information of the device,which can be calculated through the inertial measurement unit.According to the information of anchor location,the position estimation of the device and terminal heading,an error compensation model based on the mutual feedback of shadowing error and position estimation is established to achieve accurate compensation of the error caused by body shadowing and to improve the robustness of positioning.The experimental results show that the proposed method outperforms the existing quartic polynomial fitting-based compensation method in external accordant accuracy by about 22.7%.Meanwhile,the internal accordant accuracy under different body-shadowing states is improved by about 44%after the error compensation.3.In response to the accurate fusion problem for multi positioning features in device localization,the state-of-the-art of the performance for current filter-based fusion methods is firstly reviewed.A multi-level fusion framework for positioning features is designed,and the positioning information from multi-sources is flexibly integrated and dynamically fused based on the belief propagation method and the particle filter(PF).Meanwhile,aiming at the optimization problem of the importance density functions in PF,the method to estimate the statistical features of the terminal locations is then proposed based on the factor graph,which could be a guidance to build the importance density function for the PF and to calculate the weights of the particles.High accuracy and high availability localization could be achieved based on the fusion framework proposed in this paper,and the simulation results under the same data resources show that the proposed fusion method improves the positioning accuracy by about 31.1%and 16.7%when comparing with the fusion methods based on the extended Kalman filter(EKF)and the PF,respectively,with small increase of the computational complexity.4.The prototype of intelligent mobile device and the experimental environments for the evaluation of the MSIFP are established,and the methods proposed in this paper are integrated in the fusion framework and evaluated through experiments in dense cluttered indoor environments.The experimental results show that the positioning accuracy of the proposed fusion method is improved by about 24.4%and 17.3%when comparing with the fusion methods based on EKF and PF,respectively,and the positioning accuracy has reached 0.62m(1?). |
PDF Full Text Request