An Improved Robust Cubature Kalman Filter Algorithm Based on Gaussian Process

The Kalman filter algorithm serves as the fundamental component of strapdown inertial integrated navigation systems. However, when deployed in complex operational environments, the system model exhibits strong nonlinear characteristics and demonstrates inadequate robustness. To address these challenges, this study employs Gaussian process theory as a computational tool to investigate the robustness enhancement of the cubature Kalman filter (CKF). An improved robust cubature Kalman filter algorithm based on Gaussian process (GP-RCKF) is proposed. This innovative approach addresses system nonlinearity through deterministic sampling strategies, establishes a Gaussian process regression-based pseudo-observation robust mechanism, and mitigates measurement interference on system performance. Experimental validation demonstrates that the proposed methodology significantly enhances both steady-state precision and robustness in strapdown inertial integrated navigation systems, while quantitatively verifying the algorithm effectiveness through comprehensive performance metrics.