Student's t-Kernel-Based Maximum Correntropy Kalman Filter

Area of Science:

• Engineering
• Signal Processing
• Control Systems

Background:

• State estimation is crucial across many fields, with the Kalman filter being a standard method.
• The Kalman filter's reliance on mean square error limits its effectiveness with non-Gaussian noise and outliers.
• Non-Gaussian noise and outliers are common in engineering, degrading Kalman filter performance.

Purpose of the Study:

• To propose a novel filter for robust state estimation in non-Gaussian noise environments.
• To enhance the accuracy and reliability of state estimation where traditional Kalman filters fail.
• To introduce the Student's t kernel-based maximum correntropy Kalman filter.

Main Methods:

• Development of a novel Student's t kernel-based maximum correntropy Kalman filter.
• Analysis of the algorithm's convergence using a fixed-point iteration method.
• Comparative simulations against Kalman filter, Huber-based filter, and maximum correntropy Kalman filter.

Main Results:

• The proposed filter significantly outperforms conventional filters in non-Gaussian noise scenarios.
• Proper selection of kernel function parameters is key to the enhanced filter's performance.
• Demonstrated robustness against outliers and non-Gaussian noise characteristics.

Conclusions:

• The Student's t kernel-based maximum correntropy Kalman filter offers improved state estimation accuracy and robustness.
• This novel approach effectively addresses the limitations of the standard Kalman filter in challenging noise conditions.
• The filter's convergence is analyzed, ensuring reliable performance in practical applications.