heal.abstract |
The main purpose of the present work is to provide a specific approach on the (state) Observer problem for general classes of nonlinear systems. Observer design is a fundamental area in mathematical control theory. In general, it concerns the estimation of the internal unmeasured states of a control system using the information provided by inputs and outputs. There is a great amount of literature concerned with the notion of Observability and Observer design and, specifically for nonlinear systems, a variety of mathematical techniques has been developed. This thesis aims at constructing nonlinear Observers with hybrid performance that estimate the unmeasured state components in finite time, namely, dead-beat Observers.
A previous work in hybrid dead-beat Observers has already been done and it is analytically presented in Chapter 2. In the present thesis, these Observers are extended for general classes of nonlinear systems and these results are presented in Chapters 3 and 4. More specifically, the systems approached in Chapter 2 are cited as linear in the unmeasured state components. Systems we extend to in Chapter 3 possess a linear and an additional nonlinear in the unmeasured state part, while their output has a linear in the unmeasured state performance. In Chapter 4, the aforementioned systems are extended furthermore and include systems whose first part is quasi-linear and also possess an additional full nonlinearity, while their measured output is quasi-linear in the unknown state.
In Chapter 2, the results rely on a strong observability hypothesis which is derived from an observability Gramian invertibility. The corresponding Observer system which is dead-beat, exploits the information acquired by the inputs and outputs which are fed as delayed signals. Also, a robustness result, concerning the proposed Observer, is presented. The finite time estimation of the unknown states in Chapter 3 and 4 relies firstly on a Persistence of Excitation hypothesis for their linear and quasi-linear parts respectively and secondly on appropriate inequalities which provide conditions under which the unknown state can be reconstructed. The latter hypotheses use tools from Nonlinear Analysis, specifically Fixed Point Theory and these are essential to construct the unknown state component in finite time. Examples which illustrate the implementation of the techniques are included in Chapters 3 and 4. All mathematical tools that are utilized throughout the thesis, along with some preliminaries from Mathematical Control Theory, are extensively presented in Chapter 1. |
en |