A Kalman Filter Approach to Model-error Control Synthesis

Book excerpt: This thesis introduces two new techniques by which the model-error control synthesis approach can be implemented on a nonlinear system. The first method utilizes two separate extended Kalman filters. Among the two filters, one is strictly used for noise filtration/state estimation and the other is used for model error prediction. The second scheme exploits a single extended Kalman filter for the simultaneous estimation of the system states and the model error. The simulation results indicate that the new model-error control synthesis approaches are extremely effective in providing closed-loop robustness in the presence of noisy measurement signal. Finally, a sensitivity of the controlled closed-loop system stability with respect to the process noise covariance is presented.