Generalized M-Estimation-Based Framework for Robust Guidance Information Extraction

Area of Science:

• Control Systems Engineering
• Signal Processing
• Robotics

Background:

• State estimation in guidance systems is challenged by non-Gaussian noise.
• Existing methods struggle with unstable noise, leading to accuracy loss and filter divergence.
• Optimal kernel width selection is difficult with statistically undefined noise.

Purpose of the Study:

• To develop a robust framework for state estimation under non-Gaussian noise.
• To enhance the accuracy and reliability of guidance information extraction.
• To address limitations in kernel width selection and filter divergence.

Main Methods:

• Linearizing nonlinear models using statistical linear regression.
• Integrating generalized M-estimation with the Information-theoretic Maximum Correntropy Criterion Filter (IMCCF).
• Employing Singular Value Decomposition (SVD) for numerical stability and the DCS kernel function for severe non-Gaussian noise.

Main Results:

• The proposed framework demonstrates precision in Gaussian noise.
• High accuracy is maintained under significant non-Gaussian noise, proving robustness.
• Improvements in numerical stability and adaptive noise suppression enhance system reliability.

Conclusions:

• The developed algorithm effectively handles non-Gaussian noise in guidance systems.
• It offers enhanced robustness, accuracy, and reliability across diverse interference scenarios.
• This work benefits guidance system designers and filtering researchers focused on robust estimation.