Robust Evolving Granular Feedback Linearization

Abstract

This paper develops an adaptive feedback linearization approach to control nonlinear systems under model mismatch conditions. The approach uses the participatory learning modeling algorithm to estimate the nonlinearities from data streams online, and the certainty equivalence principle to compute the control signal. Simulation experiments with the classic surge tank level control benchmark show that evolving robust granular feedback linearization outperforms exact feedback linearization.