In many dynamical systems, the variables that can be directly measured with physical sensors are not necessarily the variables of interest for monitoring, diagnosis, or control. Software-based reconstruction of unmeasured states and parameters can therefore provide a valuable complement to physical measurements. An observer can be viewed as a software sensor: it consists of a dynamical model of the system coupled with measured outputs through suitable correction terms, with the objective of making the observer synchronize asymptotically with the actual system. A specific approach was proposed by Hoang et al, to construct observers for a particular class of nonlinear systems. It addresses systems for which the error dynamics can be linearized after a regular or singular time-scaling. This method relies on a switching observer that combines different observer dynamics depending on the operating region of the system. This approach makes it possible to construct observers for systems for which standard observer-design techniques may not be directly applicable. The objective of this internship is to investigate extensions of this methodology and to assess its applicability to more complex nonlinear systems, in particular to systems with unknown parameters.

• Niveaux d’étude
• Stage Bac+5