Nonlinear observer design using Automatic Differentiation
by
Klaus Röbenack
Institut für Regelungs- und Steuerungstheorie, TU Dresden, Germany
Coauthors: Kurt J. Reinschke

During the last years, many algorithms for controller and observer design for state-space systems x'=f(x)+g(x) u and y=h(x) have been developed. Some design methods are based on differential geometric or differential algebraic concepts. We show how the use of Automatic Differentiation avoids these cumbersome computations.

State feedback control requires the value of the state x. An observer can provide an estimate of this state. As for general nonlinear systems, observer design is still an actual research problem. In case of nonlinear systems with an affine input, a few observer design algorithms are known. The application to non-trivial systems is limited due to a burden of symbolic computations involved. The authors propose an observer design method based on Automatic Differentiation. The interrelations between the applicability of the algorithm and known observability conditions are discussed.

http://homepages.feis.herts.ac.uk/~comqun/AD2000/Ext_Abstracts/roebenack.ps

Date received: December 3, 1999