| With the improvement of people’s daily car travel demand,navigation and positioning technology is getting more and more attention.Navigation system can effectively assist car travel and solve the travel problems in unfamiliar environments.Therefore,it has developed rapidly in recent years.Common navigation systems include inertial navigation system（INS）and satellite navigation,or effectively combine them.Before the integrated navigation system starts working,it is necessary to conduct initial alignment to determine the reference navigation coordinate system,it is also the basis to vehicle integrated navigation system.Quick and accurate is the most great characteristics.When the accuracy of inertial elements in the integrated navigation system is not high,the system is greatly disturbed,or when the vehicle enters the communication interruption area,the GPS navigation information is obviously attenuated,and sins becomes the only available navigation system at this stage,there will be an obvious large misalignment angle in the initial alignment process,and the error model will also become a typical nonlinear model;Similar to other nonlinear filtering algorithms,when EKF algorithm is used to deal with the nonlinear initial alignment error model under large misalignment angle,the alignment effect is not ideal at some times,and even the alignment error can not converge for a time.Therefore,by analyzing the nonlinear factors existing in the initial alignment process,this paper hopes to find a method to improve the system performance from the perspective of the stability and observability of the error model in EKF algorithm,so as to improve the effect of initial alignment.Firstly,according to the working principle of integrated navigation system and the transformation of coordinate system commonly used in navigation,the attitude matrix is obtained,and then according to the angular motion relationship between misalignment angle and navigation coordinate system and the working principle of accelerometer,the nonlinear initial alignment error model is obtained,and the influence of the nonlinear part of the model on the initial alignment process is analyzed.Secondly,compared to a variety of nonlinear filtering algorithms,the research object is the most commonly used EKF algorithm.Combined with the research method of nonlinear system performance,the method to study the performance of error model in EKF algorithm is determined.The observability is analyzed by linearization method and rank criterion,and the observable error model is obtained by simplification;In terms of the stability of the system,based on the stability conditions proposed by Reif,the stability of the error model is analyzed in detail from four aspects:Jacobian matrix,linearization error,error covariance matrix and the boundary of initial error and noise term.It is explained theoretically that the boundary requirements of initial error and noise variance are too strict to meet in practice.To some extent,it explains the reasons for the poor alignment effect of EKF algorithm in some cases;Combined with relevant numerical simulation,the above analysis conclusions are further verified.Thirdly,in order to improve the observability and stability of the nonlinear error model in the algorithm,improve the alignment effect,and retain the nonlinear factors in the model,the EKF-A algorithm is proposed,and the differences in observability,stability and alignment effect between the algorithm and the traditional EKF algorithm are compared.The simulation results show that by retaining the nonlinear factors,Under the large initial error and noise variance,the observability and stability of the model in the algorithm are significantly improved,and the alignment effect is also improved.Finally,the specific influence of nonlinear factors on the alignment process is further analyzed from the two aspects of error sources and the contribution of initial error covariance.The error sources in the model will affect the alignment effect through nonlinear factors.The simulation results further illustrate the necessity of retaining nonlinear factors;The value of the initial error covariance P0will affect the alignment speed of the alignment process.Properly adjusting the size of P0is conducive to improve the alignment effect;Combined with the Kalman gain K,the method of adjusting the noise covariance parameters Q and R is given,that is,reducing Q will reduce the Kalman gain K and the sky misalignment angle,make the alignment process more stable;Decreasing R will increase the misalignment angle between K and sky misalignment angle error alignment speed is faster,but the error may increase.The above analysis is verified by experiments. |
