A PCA-based Kalman estimation approach for system with colored measurement noise

In this article, principal component analysis (PCA) is applied to improve Kalman state estimator performance in the presence of colored measurement noise without extending the state estimator dimension. Unlike the common methods the proposed PCA-based Kalman state estimator doesn't use the information of noise dynamics. First, measurements of the Sensors are entered to the PCA block. The new measurement data and the previous ones, stored in PCA buffer, merged and processed. The PCA output will be noiseless data that increase the accuracy of the Kalman state estimator. An illustrative example is simulated for comparisons of standard Kalman estimator, state augmented Kalman estimator and the PCA based Kalman estimator. Finally the simulations demonstrate the significant improvement in accuracy and performance of state estimation using the proposed method.