Observer design with guaranteed bound for LPV systems

This paper deals with state observationof discrete time LPV systems in the special situation when theparameters are not exactly known (estimated with a finite accuracyor affected by noise or bounded disturbances during theirmeasurement). It was shown that despite of theresulting mismatch between the true parameters and the availableones, the state reconstruction error boundedness can be guaranteedand an explicit but possibly conservative can be derived. Theobjective here is to propose a procedure in order to improve thegap between this bound and the actual one. The main resultconsists in a convex optimization problem which allows to selectamong all possible polytopic observers the one which minimizes theobservation error bound in the steady state. The same example is used to illustrate the improvement providedby the proposeddesign procedure.