A Switched Adaptive Observer Design Without Persistence Of Excitation

Abstract

An adaptive observer is proposed using a switched parameter estimator which requires a mild initial excitation (IE) condition on the regressor. Unlike traditional adaptive observer designs which need a stringent persistence of excitation (PE) condition on the regressor for convergence of states and the unknown system parameters, IE is less restrictive. Unlike PE, the IE condition is online verifiable, since it does not rely on the future behavior of the signal. A two-layer low-pass filter design is proposed, where the first-layer provides a parametrized observer framework, while the latter layer along with the switching mechanism averts the need of strict PE condition. Reportedly this is the first work where the need of the PE condition is obviated for the design of adaptive observer.