Sequential processing and performance optimization in nonlinear state estimation

Abstract

We propose a framework for designing observers for noisy nonlinear systems with global convergence properties and performing robustness and noise sensitivity. Our state observer is the result of the combination of a state norm estimator with a bank of Kalman-type filters, parametrized by the state norm estimator. The state estimate is sequentially processed through the bank of filters. In general, existing nonlinear state observers are responsible for estimation errors which are sensitive to model uncertainties and measurement noise, depending on the initial state conditions. Each Kalman-type filter of the bank contributes to improve the estimation error performances to a certain degree in terms of sensitivity with respect to noise and initial state conditions. A sequential processing algorithm for performance optimization is given and simulations show the effectiveness of these sequential filters.