Research On Low Cost Single Frequency RTK/MEMS-IMU/VIO Fusion Position

With the demand for advanced precision navigation and positioning systems in low-speed unmanned consumer markets such as smart agricultural machinery,low-speed unmanned vehicles,and micro drones,there is a practical need to develop a low-cost,high-precision real-time navigation and positioning system.In the current consumer-grade positioning market,most of them use the GNSS single-point positioning mode and consumer-grade MEMS-IMU combined positioning,which cannot meet the needs of high-precision positioning in the low-speed unmanned consumer market.This paper studies low-cost single-frequency RTK/MEMS-IMU/VIO fusion positioning system,based on consumer-grade MEMS-IMU,single-frequency RTK and VIO and other multi-source sensor data,on the embedded platform STM32 through extended Kalman filtering to achieve high accuracy Real-time positioning system,and conducted an experimental analysis of the system.The main research contents of this article are as follows:(1)The mathematical models of consumer-grade sensors such as single-frequency RTK,MEMS-IMU,barometer and magnetic sensor are described.In view of the problem that the performance of consumer-grade GNSS antennas affects the performance of single-frequency RTK,the consumer-grade GNSS antennas on the market are analyzed by the two indicators of pseudo-range residual and ambiguity initialization time;It can make the single-frequency RTK reach the best state and provide a good data source for multi-sensor fusion positioning.Analyze the error noise of MEMS-IMU statically through Allan variance,lay the foundation for its noise modeling,data fusion filtering;calculate the variance of the barometer and magnetic sensor data to determine its measurement error.(2)Low-cost multi-sensor real-time fusion positioning system is designed.Taking into account the low performance of a single consumer-level sensor and unable to provide stable positioning,this paper studies the low-cost RTK,MEMS-IMU and VIO fusion positioning algorithm,and determines the initialization,system noise,and measurement noise of the extended Kalman filter for consumer-level multi-source sensors.With regard to other parameters,the problems of time synchronization of multiple sensors and robust processing of outliers are considered,and a hardware platform is designed to run the multi-sensor fusion positioning algorithm in real time.(3)In order to verify the low-cost single-frequency RTK / MEMS-IMU fusion positioning performance,this paper builds an experimental platform to analyze the dynamic and static positioning of the single-frequency RTK / MEMS-IMU combined system.Based on the fusion positioning algorithm designed in this paper,Test and analyze the data of static scene,dynamic scene without occlusion and dynamic scene with occlusion.The experimental results show that: in static scenes,the combined positioning accuracy is better than 1.5cm,which is equivalent to the dual-frequency commercial receiver accuracy;In the dynamic unobstructed experiment,the fusion positioning accuracy is better than 5cm;In the dynamic covered experiment,the combined The point accuracy is 6.8cm.In the full-occlusion environment,it can't provide centimeter-level positioning accuracy.(4)In view of the problem that the low-cost RTK/IMU fusion system cannot provide high-precision navigation and positioning in the case of long-term interference or occlusion of GNSS signals,this paper fuses VIO data on the basis of RTK/MEMS-IMU to perform navigation performance analysis and GNSS simulation interrupt test.First,real-time fusion positioning of RTK/MEMS-IMU/VIO data,analysis of filter information and filtering accuracy,and the result of fusion positioning is reliable;The RTK data in the case of turning in the trajectory is artificially interrupted,and the fusion positioning experiment is carried out.The experimental results show that:In the absence of GNSS signals,RTK / IMU generates large errors in the recursion process,and the trajectory is not smooth,which cannot provide good navigation information;In the fusion positioning fused with VIO data,the trajectory of the positioning result is smooth,and the error is relatively small.The accuracy in the north direction is 0.09 m,in the east direction is 0.15 m,and in the vertical direction is 0.2m.