Forward-Backward Set-Membership State Estimator Based on Interval Analysis

International audienceInterval computation is used to design a set- membership state estimator for uncertain discrete-time linear systems. Based on the observability property, an output set-version technique is applied to discard all the parts of the predicted state enclosure that are inconsistent with systemoutput measurements over a finite time-horizon. Thanks to this correction procedure the convergence of the width of the estimated state enclosure can be established. Through a numerical example, the performance of the proposed set-membership state estimator are illustrated and compared to those of an ellipsoidal state estimator