Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
564782 | Digital Signal Processing | 2013 | 10 Pages |
A novel Gaussian state estimator named Chebyshev polynomial Kalman filter is proposed that exploits the exact and closed-form calculation of posterior moments for polynomial nonlinearities. An arbitrary nonlinear system is at first approximated via a Chebyshev polynomial series. By exploiting special properties of the Chebyshev polynomials, exact expressions for mean and variance are then provided in computationally efficient vector-matrix notation for prediction and measurement update. Approximation and state estimation are performed in a black-box fashion without the need of manual operation or manual inspection. The superior performance of the Chebyshev polynomial Kalman filter compared to state-of-the-art Gaussian estimators is demonstrated by means of numerical simulations and a real-world application.