Research On Modeling And Algorithms Of GNSS/INS Positioning & Attitude Determination

Compared with inertial navigation system(INS),the biggest advantage of the GNSS attitude measurement system is that the error does not accumulate over time and the precision is relatively stable.It has been applied to many fields such as land,ocean,sky and space.The GNSS attitude measurement system and INS complement each other,forming a GNSS/INS integrated positioning and attitude determination system can provide position,velocity,and attitude information of higher-frequency and higher-accuracy.This dissertation focuses on GNSS attitude measurement and GNSS/INS integrated positioning and attitude determination.It mainly covers the GNSS attitude measurement methods and development of attitude parameter determination methods,the filtering model construction of GNSS/INS integrated positioning and attitude determination,and the adaptive filtering algorithms of GNSS/INS integrated positioning and attitude determination.In these three parts,the main contents and innovations are as follows:1.The principle of GNSS single-antenna attitude measurement is deduced in detail,and the accuracy of single antenna attitude measurement is analyzed based on the measured data.The relationship between attitude accuracy and carrier speed is discussed.The results show that the yaw accuracy can reach within 1° when the carrier speed is more than 5m/s.This method is not suitable for the low speed state.The smaller the speed is,the higher the error level is,and it is completely unavailable at static state.2.The method for attitude parameters solution in multi-antenna attitude measurement is researched.Based on the method of least squares estimation of attitude parameters,different linearization methods are discussed.The observation equations are linearized by the additive Euler angle error,the additive quaternion error,the multiplicative Euler angle error,and the multiplicative quaternion error respectively.Four attitude parameter estimation methods are deduced and analyzed.The results show that the attitude precision of least squares method is more than 60% higher than that of the direct method.Especially,the least squares method based on multiplicative error is the best among all methods.3.The application of the total least squares in attitude parameter estimation is discussed,and the original attitude parameter estimation method is extended.For the least squares attitude parameter estimation method,both the coefficient matrix and the observation value contain errors,and the coefficient matrix contains both random and fixed elements.To solve this problem,a total least squares method of attitude parameter estimation based on Partial errors-in-variables(Partial-EIV)model is proposed.Through experimental verification,the results of the proposed method are correct,and its solution accuracy is significantly better than that of conventional least squares method.It is theoretically more rigorous.4.The navigation performance of GNSS/INS loosely-integrated and tightly-integrated are analyzed.The results show that the tightly-integrated possesses higher accuracy,stability and reliability than loosely-integrated.The filtering models of single antenna GPS/gyroscope integrated attitude determination and GNSS/INS completely integrated attitude determination are deduced in detail.Significantly,the attitude measurement effect of the completely-integrated at dynamic and static is verified and analyzed.The experimental results show that the attitude accuracy of the completely-integrated model is better than that of the tightly-integrated model,and the yaw accuracy is significantly improved.When the carrier changes from dynamic to static,the yaw of the tightly-integrated model is slowly diverging,while the completely-integrated model can overcome the problem.5.The robust adaptive Kalman filtering algorithm of GNSS/INS tightly-integrated is investigated.It's found that the conventional GPS/INS tightly-integrated robust adaptive filtering is not suitable for less than 4 satellites,and the adaptive factor constructed by predicted residuals can be easily polluted by observation abnormity.For these limitations,two improvements are proposed.Firstly,Two-step filtering is utilized.The first step's conventional EKF filtering residuals are utilized to construct the second step's gross error discriminant.Secondly,in the second step,when constructing the adaptive factor by predicted residuals,the residuals and its covariance of abnormal observations are removed to weaken its harmful effect on adaptive factor.The experiment results show that the improved method can correctly modified the gross errors and dynamic model abnormities when the satellite number is less than 4 and the observed values are abnormal,which greatly improves the reliability and stability of the filtering.6.The analysis results of the GNSS/INS completely integrated attitude measurement shows that there is still a problem that the yaw accuracy deteriorates when the carrier is steered at a large angle.The reason is that the dynamic model abnormities are increased at large-angle steering.The study found that under the premise of reliable external attitude observation,the yaw observations of the completely-integrated can directly reflect the state abnormities of the yaw parameter.Based on this feature and the adaptive filtering theory of the classification factor,a yaw angle adaptive filtering algorithm for completely-integrated is proposed.The experimental results show that this method can effectively weaken the influence of the dynamic model abnormities when the carrier is turning at a large angle,which improves the yaw precision.