Analysis Method of Strapdown Inertial Navigation Error Distribution Based on Covariance Matrix Decomposition

TL;DR

This paper introduces a covariance matrix decomposition method to analyze and quantify the contribution of various error sources in Strapdown Inertial Navigation Systems, improving understanding and optimization of navigation accuracy.

Contribution

The paper presents a novel error distribution analysis method based on a 34-dimension SINS error model that considers error interactions and is computationally efficient.

Findings

The method aligns with error propagation laws in simulations.

It effectively decomposes navigation errors into source contributions.

It outperforms Monte-Carlo methods in efficiency and comprehensiveness.

Abstract

Error distribution analysis is an important assistant technology for the research of SINS(Strapdown Inertial Navigation System). Error distribution result can provide the contribution of different errors to final navigation error, which is helpful for modifying and optimizing SINS. To realize decomposing the navigation error into parts that caused by each error source, the SINS error state space model is established and covariance matrix is decomposed according to error sources. The proposed error distribution analysis method based on 34-dimension SINS error model can quantitatively analyze the contribution to the end navigation error of initial errors, IMU(Inertial Measurement Unit) bias, IMU scale factor errors, mounting errors of gyroscopes and accelerometers, and IMU stochastic errors. The simulations in static condition and single axis rotation condition indict that the distribution result of proposed analysis method accords with the law of error propagation. After trajectory determined, the corresponding error distribution result will be calculated with the proposed method. Compared with the Monte-Carlo method and other method based on covariance matrix, the proposed method uses more complete error model, considers the interaction effect of error sources and can be easily realized with less computation.