Conference paper

Proceedings of the 9th International PhD Workshop on Systems and Control

Jožef Stefan Institute

Ljubljana

2008

978-961-264-003-3

Ing. Lenka Pavelková, Ph.D. (ÚTIA - Oddělení adaptivních systémů)

state model, uniform innovations, state filtration, paramete

The linear state-space model with uniform innovations (LU model) proposed in previous author's work is extended here. The states and parameters of LU model are estimated under hard physical bounds. The estimation of the innovation boundaries is also included. Maximum a posteriori probability estimation reduces to the linear programming. The on-line estimation is running within the sliding window. Compared to the original model, we consider that model matrices can be time-variant. Also, offset terms are included. We present the problem of the joint parameter and state estimation, i.e., the state filtering with unknown model matrices. The ambiguity in the state estimates can be substantially decreased by the partial knowledge of some entries in the model matrices. The simple example illustrates this approach.