【作者】 陈伯钧； 修春娣； 李金昆；

【Author】 CHEN Bojun;XIU Chundi;LI Jinkun;School of Electronic Information Engineering,Beihang University;

【机构】 北京航空航天大学电子信息工程学院；

【摘要】 针对室内行人定位需求,研究基于Wi Fi、地磁和行人航位推算(pedestrian dead reckoning,PDR)的多源融合定位算法。首先利用Wi Fi约束地磁进行匹配定位,并进一步基于扩展卡尔曼滤波(extended Kalman filter,EKF)、反向传播(back propagation,BP)神经网络、因子图优化算法,实现匹配定位结果和PDR航位推算结果的融合。为了验证融合定位算法的有效性,利用真实场景中的实测数据对三种融合算法的定位性能进行仿真分析。结果表明,基于BP神经网络的融合定位算法适用于高精度定位场景,而综合考虑定位精度、算法耗时和复杂度,基于因子图优化的融合定位算法更为实用。更多还原

【Abstract】 Aimed to the needs of indoor pedestrian positioning,this paper focuses on multi-source fusion positioning algorithms based on wireless fidelity(WiFi),geomagnetism,and pedestrian dead reckoning(PDR).Firstly,WiFi is used to constrain the geomagnetic field for improving the positioning accuracy preliminarily.Secondly,based on three kinds of multi-source fusion location algorithms:extended Kalman filter(EKF),back propagation(BP) neural network,and factor graph optimization,the positioning results of PDR and WiFiconstrained geomagnetic are fused.Lastly,to verify the effectiveness of the fusion positioning algorithm,simulate and analyze the positioning performance of three fusion algorithms using measured data in real scenarios.The results show that the fusion positioning algorithm based on BP neural network is more suitable for highprecision localization.Considering the positioning accuracy,time consuming and complexity,the fusion positioning method based on factor graph optimization is preferred.更多还原