H control for non-linear stochastic systems: The output-feedback case

N. Berman, U. Shaked

Research output: Contribution to journalArticlepeer-review

Abstract

The H output-feedback control problem for non-linear stochastic systems is considered. A solution for a large class of non-linear stochastic systems is introduced (including non-linear diffusion systems as a subclass). This solution is based on a bounded real lemma for non-linear stochastic systems that was previously established via a stochastic dissipativity concept. The theory yields sufficient conditions for the closed-loop system to possess a prescribed L2-gain bound in terms of two Hamilton Jacobi inequalities: one that is associated with the state feedback part of the problem is n-dimensional (where n is the underlying system's state dimension) and the other inequality that stems from the estimation part is 2n-dimensional. Both stationary and non-stationary systems are considered. Stability of the closed-loop system is established, both in the mean-square and the in-probability senses. As the solution to the Hamilton Jacobi inequalities may, in general, lead to a non-realisable state estimator, a modification of the associated 2n-dimensional Hamilton Jacobi inequality is made in order to circumvent this realisation problem, while preserving the system's L2-gain bound. For time-invariant systems, the problem of robust output-feedback is considered in the case of norm-bounded uncertainties. A solution is then derived in terms of linear state-dependent matrix inequalities.

Original languageEnglish
Pages (from-to)1733-1746
Number of pages14
JournalInternational Journal of Control
Volume81
Issue number11
DOIs
StatePublished - Nov 2008

Keywords

  • Disturbance attenuation
  • H
  • Linear state-dependent noise
  • Matrix inequalities
  • Nonlinear systems
  • Stochastic stability
  • Stochastic systems

Fingerprint

Dive into the research topics of 'H control for non-linear stochastic systems: The output-feedback case'. Together they form a unique fingerprint.

Cite this