Noise-Correlated Two-Stage Cubature Kalman Filtering Estimation Algorithm

Existing two-stage filtering algorithms in the literature assume that the system is nonlinear and Gaussian, with independent noise, meaning the noise in the state equation and measurement equation are uncorrelated and both follow Gaussian white noise distributions. However, in practical applications, noise correlations are common, and traditional computational methods that ignore these correlations inevitably lead to reduced estimation accuracy. This paper proposes a Noise-Correlated Two-Stage Cubature Kalman Filtering Algorithm (TSCKF-CN) based on model transformation. The algorithm introduces a coefficient O k to transform the model from a noise-correlated system to a noise-independent system. It then employs the noise from the transformed model in the recursive computation of the two-stage filter to achieve a noise-correlated TSCKF estimator. Simulation results from a pure azimuth tracking system demonstrate that this method, by accounting for noise correlation, achieves better tracking accuracy than methods that neglect noise correlations, leading to improved tracking performance.