# A Scheme for System Error Calibration and Compensation of the Initial State of MEMS Inertial Navigation

## TL;DR

This paper introduces a cost-effective method to improve the accuracy of MEMS inertial navigation systems by calibrating and compensating for initial errors without expensive equipment.

## Contribution

A novel twelve-position dual-axis calibration method that achieves high accuracy and reduces reliance on three-axis turntables.

## Key findings

- Installation error calibration accuracy reaches 0.03°, a 25% improvement over traditional methods.
- Navigation error is reduced by 90% within one hour, with initial error ≤ 1 μrad.
- The method enables multi-error calibration while addressing the cost–accuracy trade-off.

## Abstract

Aiming at the challenge of balancing the accuracy and cost of the initial state calibration of traditional MEMS inertial navigation systems, as well as the current situation of the lack of high-precision three-axis turntables in engineering practice, this paper proposes a practical and innovative systematic error calibration and compensation scheme, which effectively suppresses the deterministic errors of MEMS-INS and enhances its applicability in high-precision and long-duration tasks. By analyzing the coordinate transformation characteristics of the MEMS-INS solution process under small-angle disturbances, a deterministic error model based on the device’s zero bias, scale factor, and cross-coupling errors is constructed. A twelve-position dual-axis calibration method, combined with a high-precision orthogonal fixture, is designed to excite errors on a dual-axis turntable, converting originally unobservable error terms into observable periodic signals. Experimental results show that the installation error calibration accuracy reaches 0.03°, an improvement of about 25% compared to the traditional dual-axis method, breaking through the limitations of dual-axis turntables in cross-coupling error calibration, achieving an initial error ≤ 1 μrad, and reducing the navigation error by 90% within one hour. This method eliminates reliance on expensive three-axis turntables while enabling multi-error calibration, addressing the cost–accuracy trade-off in engineering applications.