Research On Continuous Tumbling Self- Calibration Test Design Method For Inertial Navigation Platform System

This paper analyzes the influence of the error of the inertial navigation platform on the accuracy during the actual use of the inertial platform.Through continuous rolling self-calibration method,it compensates during the operation.In the continuous rolling self-calibration process,the continuously rolling trajectory has a great influence on the test results.Using the Particle Swarm Optimization(PSO)global optimization algorithm to find the continuous rolling self-calibration optimal rotation trajectory,all error information of the system can be obtained.The system-level state equations and output equations of the inertial navigation system are established.Based on the platform space and the computational space alignment error angle ? angle method,an error model for the continuous roll selfcalibration test of an inertial navigation platform system is established.According to the error model of accelerometer and gyroscope,the transformation between coordinate spaces establishes the system's state equation and observation equation.The observability of the system is studied,and the influence of the determinant of the information matrix and the rotation orbit of the platform on the observability is given.Use the D-optimal design to find the optimal rotation trajectory corresponding to the maximum value of the information determinant.Through the D-optimal experimental design method,the search for the optimal rotation trajectory is converted into the maximum value of the determinant M of the information matrix.The PSO global optimization algorithm is used to solve the maximum value of the determinant M of the information matrix with nonlinear constraints.This paper focuses on the analysis of the nonlinear constraints when PSO global optimization algorithm solves this problem.By analyzing the general flow of using PSO algorithm to obtain the unconstrained maximum value,the key steps of nonlinear constraint processing are added on this basis.The nonlinear constraints in the model make the PSO algorithm put forward new requirements for the initialization of the particle swarm.Before the algorithm runs,it is necessary to first adjust the particle swarm to make it conform to the nonlinear constraints instead of randomly initializing the particle swarm.The penalty function method and the direct method are used to deal with nonlinear constraints.The absorption boundary method,reflection boundary method,and invisible boundary method in the direct method have all achieved very good experimental results.Selecting parameters and tracking the iterative process of the algorithm illustrates the effectiveness of the PSO algorithm in solving the model.The influence of the parameters of the PSO on the test accuracy was studied.Different parameters correspond to different test results.The influence of the inertia weight coefficient W on the algorithm is analyzed,and the range of the optimal inertia weight coefficient is given.The influence of cognitive factor 1c and social factor 2c on the algorithm is analyzed,and their optimal parameter combinations are given.Finally,the most suitable combination of parameters is selected to obtain a continuous rolling self-calibration optimal rotation trajectory.Aiming at the comparison of the effects of the optimization algorithm,the advantages and disadvantages of the gradient-based global optimization method and the random optimization method are analyzed.The deficiencies of the gradient-based optimization method demonstrate the ability of stochastic optimization to handle multivariable,high-dimensional,and nonlinear models.Comparing the PSO algorithm used in this paper with the classical gradient optimization method,the advantages of the PSO intelligent algorithm are demonstrated through simulation.Compared with simulated annealing algorithm and genetic algorithm,the simulation shows that the algorithm takes less time and has better results.