RODM-Alignment: An Inertial Alignment Algorithm for Sliding Window Reverse Optimization Delayed Marginalization

Jincheng Peng

doi:10.1109/TIM.2024.3438852

Return

RODM-Alignment: An Inertial Alignment Algorithm for Sliding Window Reverse Optimization Delayed Marginalization

2024-01-01

PRE

DOI：10.1109/TIM.2024.3438852

Original

Original request for help

Save

Abstract

En 中文

In the initial alignment of the strapdown inertial navigation system, it is difficult to obtain sufficient data information to make the optimized state of convergence in a short time, which cannot realize fast alignment, and the accuracy of the alignment results is poor. The dynamic sliding window is difficult to recover after marginalization, and the linearization points of the Jacobi matrix are inconsistent, leading to the zero-space variation of the information matrix. To address the above problems, we propose an inertial alignment algorithm for sliding window reverse optimization delayed marginalization (RODM) to improve the rapidity and accuracy of the alignment algorithm. The method of storing gyro and accelerometer data during the alignment process and repeatedly solving the stored inertial data using forward and reverse navigation solving and optimization algorithms (OAMs) is used as a way to extend the interval of inertial data collected in a short period of time so that there is enough inertial data to enable the convergence of the state variables of the sliding-window optimization for a short period of time to achieve fast alignment. A delayed marginalization factor is used to update the marginalized prior information, solving the problem that the linearization of the Jacobi matrix must be fixed point to improve the accuracy of the alignment algorithm. We propose an algorithm to improve the accuracy of alignment, which we have verified the rapidity and accuracy of our initial alignment system in the turntable test, where the accuracy of the alignment heading angle reaches 0.136467 mil in 3 min, providing accurate attitude information for navigation and localization.

Keywords:

Jacobian matrices

Optimization

Accuracy

Inertial navigation

Gyroscopes

Accelerometers

Earth

Factor graph optimization

initial alignment

reverse navigation

state estimation