A regularization approach to robust variable structure observer design applied to vehicle parameter and state estimation

State estimation falls under the class of applications described as inverse problems and like all such problems is often ill posed, i.e., exhibits extreme sensitivity to small errors or noise in the measurements. Sliding mode estimation techniques have been employed to construct observers for nonlinear systems, but due to their tendency to act as high gain systems may exacerbate the difficulties associated with ill conditioned systems. This paper therefore formulates the state estimation problem as a constrained optimization problem, that seeks a trade off between the information available in the measurements and the a priori knowledge about the system, usually available in the case of dynamic systems as a set of differential equations describing the evolution of the systems states in time. The paper presents a simulation example to demonstrate the noise suppression characteristics of this formulation and also presents the application of the regularized sliding observer in a truck roll warning system