An improved Kalman filter algorithm for tightly GNSS/INS integrated navigation system.

Area of Science:

• Navigation Systems Engineering
• Signal Processing
• Robust Estimation Theory

Background:

• Kalman filters utilizing Singular Value Decomposition (SVD) are effective in mitigating numerical rounding errors in various computational applications.
• Tightly coupled Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS) integration demands improved filtering performance and adaptability.
• Conventional methods struggle with fluctuating GNSS signals due to fixed noise covariance, limiting accuracy in integrated navigation.

Purpose of the Study:

• To propose an improved robust method for tightly coupled GNSS/INS navigation systems.
• To enhance filtering performance and adaptability by addressing limitations of fixed noise covariance.
• To extend the application of the information filter (IF) through its SVD form.

Main Methods:

• Developed a novel robust method that dynamically corrects noise covariance using a correction variable derived from innovation and an SVD-derived matrix.
• Constructed a new matrix via SVD on the original matrix to improve robustness.
• Derived the SVD form of the information filter (IF) to broaden its applicability.

Main Results:

• The proposed method demonstrated superior robustness and higher positioning accuracy compared to traditional SVD-based Kalman filters.
• Maximum error was reduced by 45.77% compared to the traditional SVD-based Kalman algorithm.
• The SVD-form IF algorithm achieved a 4.7% reduction in root mean squared error compared to the traditional IF algorithm.

Conclusions:

• The proposed robust method significantly improves positioning accuracy and robustness in tightly coupled GNSS/INS navigation.
• Dynamic noise covariance correction via SVD offers better adaptability to signal fluctuations.
• The SVD-based IF provides a more accurate and reliable navigation solution, validated through simulations and experiments.