A non-linear filter with higher-order weighting functions via invariant imbedding

Publisher Website
Google Scholar
Abstract
The problem considered is the sequential estimation of states and parameters in noisy non-linear dynamical systems. The class of systems considered are those in which the dynamical behaviour is described by an ordinary differential equation. No statistical assumptions are required concerning the nature of the unknown inputs to the system or the measurement errors on the output. The equations of the estimator is derived by a least squares criterion and the invariant imbedding approach. The new feature of the algorithms derived lb that a non-linear filter with higher-order weighting functions (higher-order approximated optimal filter) is obtained by using the approximate method in the function space. Simulation results are presented which yield a comparison of the performance of the higher-order approximated optimal filter versus the other nonlinear filters when applied to a chemical batch reactor system. The results indicate that the proposed non-linear estimation scheme is feasible.

This publication has 12 references indexed in Scilit: