Research On Acceleration Technology Of Satellite Positioning Correction In Urban Canyons Based On FPGA

In urban canyons,propagation path of Line of Sight(LOS)satellite signal may be blocked by buildings,at this time,using the None Line of Sight(NLOS)propagation path can avoid the problem of insufficient satellites number.However,the pseudorange error caused by NLOS sight propagation path will significantly reduce the accuracy of satellite positioning.By reconstructing the NLOS satellite propagation path,the positioning error can be effectively reduced,however,the complex operation of the process can not meet the real-time of satellite positioning.Thesis studies the acceleration of NLOS propagation path reconstruction based on Field Programmable Gate Array(FPGA).The research contents include the following three aspects.Firstly,based on the simplified 3D building model of urban canyon,the University of Tokyo proposed a method to reduce the satellite positioning error of urban canyon,through reconstructing the NLOS satellite propagation pseudorange by mirror ray tracing simulation.In this method,there are many candidates for double sampling,so image ray tracing takes a long time.Due to this situation,Thesis proposes an inherited sampling plan to reduce the number of sampling points and effectively improve the correction speed.The tested results in the urban canyon environment near the campus show that this method improves the satellite positioning correction accuracy compared with the standard single point positioning,improves the correction speed compared with the double sampling method,and meets the real-time requirements.Secondly,FPGA has algorithm universality and hardware acceleration characteristics,which can effectively accelerate the image ray tracing and improve the efficiency of reconstructing NLOS propagation path.Because of the heterogeneous architecture of CPU and FPGA in the development board,the fuzzy analytic hierarchy process is proposed to evaluate and design the load distribution plan in the heterogeneous development board.In FPGA,the floating-point intellectual property core is used as the basic operation core,the block random sequence memory is reused to reduce the resource occupation,and the image ray tracing accelerator is designed and built in parallel with the pipeline to improve the computing speed.At the same time,the inherited sampling correction algorithm is running on the Linux system of the development board,and a realtime satellite positioning correction system is established.The measured results in the urban canyon environment show that the system not only improves the positioning accuracy compared with the standard single point positioning,but also improves the calculation efficiency compared with the sampling correction plan running on the workstation.Finally,the reconstruction of NLOS path by image ray tracing takes a long time.When there are many building models,it still can not meet the needs of high real-time applications.Considering that in the urban canyon,the altitude angle and azimuth angle range determine the NLOS satellite propagation path,and there is a functional relationship between the positioning point,azimuth angle and the exist range of the urban canyon NLOS propagation path at the positioning point.The BP neural network trained by the above data can quickly reconstruct the NLOS propagation path,which can replace the image ray tracing in the satellite positioning correction plan and effectively improve the calculation efficiency.The results show that the method improves the correction accuracy compared with the standard single point positioning method,reduces the amount of calculation compared with ray tracing,and greatly improves the calculation efficiency.