Experimental Validation of a Nonlinear Self-Tuning Observer

Abstract

The present work provides an experimental validation for the robust performance of a nonlinear self-tuning observer in accurately estimating the state variables of a structure with configuration-dependent natural frequencies. The observer design combines the advantages of the variable structure systems theory and the self-tuning fuzzy logic technique. The robustness of the observer enables the use of an imprecise model of the plant while the self-tuning characteristic eliminates the need for the extensive tuning of a fixed rule-based expert fuzzy inference system. The experimental set-up consists of a spherical robotic manipulator with the protruding portion of the third link from the second link considered to be deformable. The prismatic joint induces considerable variations in the natural frequencies of the flexible element. The observer has been implemented to estimate the generalized coordinates pertaining to the flexible motion of the third link. The experimental results confirm the robustness of the observer in accurately estimating the required state variables of the system in spite of significant modeling imprecision and disturbances.