Research On INS/GNSS Integrated Navigation Based On Adaptive Kalman Filter

GNSS and INS complement each other’s advantages.The combination of the two overcomes the disadvantages of incomplete information,discontinuity,and instability of a single navigation system,and has the advantages of all-weather,high precision,and full range.INS/GNSS has become a hot topic in the research and application of navigation in recent years.Kalman filter is a state optimal estimation algorithm commonly used in the fusion processing of INS/GNSS integrated navigation data.Traditional Kalman filter is based on the minimum mean square error criterion,which can achieve the optimal estimation of navigation results when the system noise characteristics are known.However,in urban mapping,autonomous driving,and other urban applications,it cannot adjust the preset state and measurement model in real-time to adapt to complex observation environments such as urban canyons,tree-lined roads,and tunnels,as well as large maneuver driving scenarios such as elevated bridges and roundabouts.This can lead to a decrease in the accuracy and reliability of INS/GNSS integrated navigation and even cause filter divergence.This paper introduces the SageHusa adaptive filter and robust filter algorithm based on the innovation vector on the basis of traditional Kalman filter and combines the ideas of forward and backward filter to improve the post-processing positioning,velocity,and orientation performance of INS/GNSS loosely coupled navigation.The main research work and achievements of this paper include the following aspects:(1)Introduction of the commonly used coordinate systems of INS/GNSS integrated navigation,derivation of the inertial navigation update algorithm,indirect coarse alignment algorithm,and error differential equation,analysis of instrument performance using the Allan variance analysis method.The results show that the Allan variance can accurately extract the relevant parameters of the instrument.The basic principles of indirect coarse alignment and Kalman filter fine alignment are explained,and the basic formulas of the two methods are given.The calculation results of the horizontal attitude and heading angle show that Kalman filter fine alignment can achieve better results than indirect coarse alignment and accelerate the convergence speed of estimation.(2)Introduction of the basic principles and processes of Kalman filter,explanation of the core formulas of Sage-Husa adaptive Kalman filter algorithm,robust Kalman filter algorithm based on the innovation vector,and forward and backward Kalman filter algorithm,derivation of the practical correction formulas of spatial lever arm error and time asynchrony error in INS/GNSS loosely coupled navigation,and construction of a 15-dimensional error state,6-dimensional velocity,and position measurement INS/GNSS loosely coupled state space model based on this.(3)Development of the PPKLC post-processing program for INS/GNSS loosely integrated navigation on the VS2019 platform.The program can perform inertial navigation indirect coarse alignment,velocity vector-based Kalman filter fine alignment,Allan variance analysis,inertial navigation data processing,and INS/GNSS data processing based on traditional and improved Kalman filter.It can output parameter estimation information such as attitude,heading,velocity,and position,as well as the error of parameter estimation,for subsequent navigation result analysis.(4)A complex route containing a variety of typical urban navigation scenes is designed,aiming to verify the five algorithms introduced in the program by using measured data.The collected data was analyzed and preprocessed.The GNSS velocity and position were calculated using the PPK algorithm,while the initial attitude and heading of INS were calculated using indirect coarse alignment and Kalman filter fine alignment.(5)PPKLC was used for post-processing of INS/GNSS loosely integrated navigation data that has been preprocessed.The experimental results show that the five introduced algorithms can effectively improve the overall performance of the integrated navigation system with the estimation accuracy of attitude angle,heading angle,threedimensional velocity and position improved by up to 58.77%,95.17%,36.61% and49.20%,respectively,compared to the traditional Kalman filter algorithm.The two adaptive Kalman filter algorithms can accelerate the convergence time of heading angle and ensure the accuracy before convergence,with the estimation accuracy improved by6.75%,18.18%,15.77% and 15.53% on average.The two kinds of adaptive Kalman filter with the idea of forward and backward filter can achieve better data processing performance,and has significant advantages in accelerating the convergence of parameter estimation,ensuring the stability of filter and improving the estimation accuracy.The estimation accuracy is improved by 57.33%,94.94%,30.21% and 40.63%on average.In terms of convergence speed,about 300 s of course angle convergence time is saved,and the accuracy is better than traditional Kalman filter and adaptive Kalman filter in trajectory calculation.