A Pseudospectral Observer for Nonlinear Systems

Abstract

We present a method for designing an observer for nonlinear systems based on Pseudospectral discretization and a moving horizon strategy. The observer has a low computational burden, fast convergence rate and an ability to handle measurement noise. Our observer can also be applied to nonlinear systems governed by deferential-algebraic equations (DAE) which is very did_cult to deal with by other designs like the unscented Kalman filter. The performance of the proposed observer is demonstrated by numerical experiments on a time-varying chaotic nonlinear system with unknown parameters and also a nonlinear circuit with singularity-induced bifurcation.