abstract
- This paper addresses the problem of observer design with unmeasurable gain scheduling functions for Quasi-Linear Parameter Varying (Quasi-LPV) models through a LPV Kalman Filter. The proposed methodology rewrites the system as dependent on the estimated scheduling vector. This requires finding a set of offline optimal filter gains that are robust against uncertainties such as perturbations and noises, while ensuring stability and convergence at each vertex by means of a polytopic formulation. An experimental application for non-concurrent water leak diagnosis using a dataset is performed to validate the proposed approach.