Parameter and state estimation of switched affine systems

Jamal Daafouz (1) , Romain Postoyan (1) , Pierre Riedinger (1)

1 CRAN - Centre de Recherche en Automatique de Nancy

We investigate the on-line parameter and state estimation of switched affine systems. We first adapt standard techniques available for affine (time-varying) systems to the considered class of systems, which leads to new linear matrix inequality-based conditions for the design of the estimator gains. We then explain how to derive decoupled estimators in the sense that the state and the parameter estimates are generated by cascaded switched differential equations, which is numerically more efficient and thus easier to implement in practice. A first solution is presented, which generates state estimates with jumps at the switching instants. To overcome this potential issue, two "bumpless" strategies are proposed. The first one relies on a persistency of excitation (PE) condition, while the the second one applies to a given class of switching rules and avoids the PE condition. Both estimation strategies ensure the global exponential convergence to zero of the estimation errors. The results are illustrated on a model describing the behavior of a DC-DC power converter.

Mots clés: Switched systems; Estimation; Observers for nonlinear systems

Domaines: Automatique / Robotique

Jamal Daafouz, Romain Postoyan, Pierre Riedinger. Parameter and state estimation of switched affine systems. 56th IEEE Conference on Decision and Control, CDC 2017, Dec 2017, Melbourne, Australia. ⟨hal-01652490⟩