| Along with the increasement of Location-Based Service(LBS),the high precise and robust positioning technique is concerned by academia and industry.A lot of emerging technologies,such as robot,intelligent drive technique and so on,which are different from traditional LBS applications,are applied in complex environment and scene.In view of only single positioning method can't satisfy demands,the technology of multi-sensor fusion is increasingly concerned and researched by scholars.In the context,this paper combines the GNSS Precise Point Positioning(PPP),atmospheric augmentation,Inertial Navigation System(INS),Visual Odometer(VO)for the research and performance analysis of the fusion positioning algorithm.Firstly,we model the tightly coupled integration of GNSS and INS based on the atmospheric augmentation PPP,deduce the mathematical model of integration algorithm.Meanwhile,utilizing three different scenes set of vehicle data to analyze the performance of tightly coupled integration.The results are: 1.The product of atmospheric augmentation is more precise than traditional prior models,whose accuracy of zenith troposphere delay is better than 4mm,and the ionosphere accuracy is less than 1.4TECU;2.In the open sky scene,introducing atmospheric augmentation can bring an obviously improvement of single-frequency PPP,which is about 45.3%,and a little improvement of dual-frequency PPP,whose vertical component improved better than horizontal component;3.In the open sky scene,atmospheric augmentation PPP/INS tightly coupled integration performance much better than PPP-only,especially with single-frequency GNSS data.Including the both atmospheric augmentation and INS can bring an improvement about 25% with dual-frequency data and 55% with single-frequency data.Both of single-frequency and dual-frequency PPP/INS integration can achieve a decimeter-level positioning;4.In complex scene,such as urban and tree-lined road,including INS can improve the performance of PPP about 30%,and then introducing atmospheric augmentation to PPP/INS still improve the positioning accuracy,which compared with PPP better than 40% improvement and achieve a meter-level positioning of single-and dual-frequency data;5.In complex scene,the positive effect of INS improve the gross error and some large error values,and atmospheric augmentation minimize systematic error.Then,based on PPP/INS tightly coupled integration,introducing the VO into the coupled integration system.On the basis of Multi-State Constraint Kalman Filter(MSCKF),deduce the mathematical model of PPP,INS and visual information fusion to establish a new fusion frame of the multi-sensor,which can balance the robust and precision.At the same time,the equipment and platform of acquiring multi-sensor data are established,with achieving spatial and time synchronization of the three sensors.After testing and analyzing the performance of multi-sensor positioning with complexscene data of tree-lined road in campus,the results mean: 1.The performance of Visual Inertial Odometry(VIO)are much better than INS-only,whose relative positioning precision is about 0.67%.Howerver,in view of a large-distance motion,its absolute positioning precise cannot put on a par with GNSS PPP;2.In the extremely complex scene,the precision of single-or dual-frequency PPP is not acceptable with a bad noise and dis-robustness.Contrast with PPP,multi-sensor fusion preforms much better whatever with single-or dual-frequency data,both of single-or dual-frequency data are able to achieve the accuracy with 2.7m in three-dimension,which better than single-and dual-frequency PPP about 43.79%,and 32.96%,respectively;3.In view of multi-sensor fusion,single-frequency data can achieve a similar preformation compared with dual-frequency,it is significance to low-cost applications.Meanwhile,with multi-sensor fusion,the result's noise is smoothed and the system's reliability and robustness are improved. |
PDF Full Text Request