Variable Bayesian-Based Maximum Correntropy Criterion Cubature Kalman Filter with Application to Target Tracking.

Area of Science:

• Signal Processing
• Control Systems
• Machine Learning

Background:

• Radar target tracking faces challenges from nonlinear dynamics, non-Gaussian noise, and sensor outliers.
• Existing robust methods struggle with empirical tuning and computational load, limiting performance.
• A need exists for adaptive, efficient, and robust filtering for complex tracking scenarios.

Purpose of the Study:

• To propose a fully adaptive and robust filtering framework for nonlinear systems.
• To address limitations of current methods in noise suppression and real-time efficiency.
• To develop a filter that eliminates manual empirical adjustments.

Main Methods:

• Introduced the variational Bayesian-based maximum correntropy criterion cubature Kalman filter (VBMCC-CKF).
• Integrated variational Bayesian inference with Cubature Kalman Filter (CKF).
• Modeled kernel size as an inverse-gamma distributed random variable for joint state and parameter optimization.

Main Results:

• VBMCC-CKF demonstrated robust performance in single and multi-target tracking under non-Gaussian noise.
• Achieved at least 14.33% reduction in average root mean square error (Avg-RMSE) for single-target tracking.
• Showcased a 40% lower optimal subpattern assignment (OSPA) distance and superior hit rates in cluttered environments.

Conclusions:

• The VBMCC-CKF framework offers a precise and adaptable solution for dynamic target tracking.
• The method achieves balanced noise suppression and real-time computational efficiency.
• It effectively overcomes the limitations of empirical tuning in conventional filters.