Experimental Design For Testing Angular Dynamic Errors Of IMU Based On Large-Scale Optimization Method

The accuracy of the navigation system is an extremely important performance indicator for measuring the pros and cons of a strapdown inertial navigation system.A series of related error models have been studied at home and abroad.The error of the inertial measurement unit(IMU)occupies a large proportion,and its dynamic error has a great influence on the actual application.However,the related research is relatively poor.This paper is aiming at the calibration trajectory optimization of IMU dynamic error coefficients,which is operated on a two-axis gimbal table.An optimal method based on Reduced Space Search Algorithm(RSSA)and the Nonlinear Conjugate Gradient(NCG)method is proposed.The Kalman filter is used to identify the error coefficients of each inertial instrument,and the estimated error covariance matrix is used to measure the calibration accuracy of the trajectory.It is a large-scale optimization problem because of its hundreds even thousands of variables.Additionally,the objective to minimum is a constrained non-convex function.A joint solution of RSSA and NCG is used to solve the problem.This joint method not only avoids the RSSA's random property which may result in losing the optimal value in a relatively small region,but also recuperates the NCG's disadvantage that it can only obtain the local minimum value.Simulations of the micromachined gyroscope IMU and the ring laser gyroscope IMU were performed.The results show that the optimization algorithm with NCG does indeed obtain better optimization results than simply using RSSA method.That is,the accuracy of the error calibration has been improved.The accuracy of the laser gyro IMU's misalignment angles,scale factor errors and the lever arm errors has been increased from 3% to 16% each.As for the micro-machined gyro IMU,the accuracy of the scale factor errors,the nonlinearity errors,the misalignment angles,and the lever arm errors also increase from 5% to 15% respectively.As for the computing time,the method with the NCG method is indeed more time-consuming than the space-reduction method,but it does not account for too much weight.It has increased by about 7.3% on the basis of the original.