Finite-horizon discrete-time robust guaranteed cost state estimation for non-linear stochastic uncertain systems

This study presents a new approach to discrete-time robust non-linear state estimation based on the use of sum quadratic constraints. The approach involves a class of state estimators that include copies on the system non-linearities in the state estimator. The non-linearities being considered are those that satisfy a certain generalised monotonicity condition. The linear part of the state estimator is synthesised using minimax LQG control theory which is closely related to H^∞ control theory and this leads to a non-linear state estimator that gives an upper bound on an estimation error cost functional.