Using the monotone property and sensors placement as basic tools to design set-membership state estimators
Résumé
In this work, we propose to combine the monotone property of dynamical systems with sensor placement techniques to design efficient set-membership state estimators. The basic idea exploited in this approach consists in decomposing the system as an interconnection of monotone subsystems where the input of each subsystem is a function of the outputs of the remainder subsystems. Thereby, the sensors placement is dictated by this monotone decomposition of the original system. Due to the new configuration of the system, one can efficiently cope with the pessimism propagation linked to set-membership computation and so one can enhance substantially the accuracy of the estimated time-varying state enclosure. Two versions of this guaranteed state estimation approach are presented. The first one is based on the principle of output injections. The second version relies on an event-triggered policy, which reduces greatly the use of measurements. In fact, with this policy, the measurements are used only when it is necessary.