An Augmented Multiple-Model Adaptive Estimation for Time-Varying Uncertain Systems

Abstract

An augmented multiple-model adaptive estimation (MMAE) algorithm is presented for a time-varying system, where the model uncertainty may occur occasionally. Generally, it is difficult for a single filter to achieve superior performance for both the certain system and the uncertain system. An algorithm that is designed for an uncertain system may yield suboptimal performance in the situation, where the model uncertainty does not occur. To cope with this problem, we propose to use two filters in parallel in a multiple-model framework. One of the filters, an augmented Kalman filter (AKF), provides estimates of uncertain parameters when the model uncertainties occur, whereas the second filter, a Kalman filter (KF), yields high precision in the absence of the uncertainties. A practical example is given in simulation to show the potential application of the presented algorithm. It indicates that the augmented MMAE is efficient to deal with the occasional model uncertainty.