Discrete-time data-driven disturbance-observer control based on fuzzy rules emulating networks

Abstract

Dynamical models of real-world systems are uncertain at the best extent, and the controller is usually based on an approximate model. Additional unavoidable effects, such as un-modelled disturbances, discretisation and quantisation, as well as measurement and input noise, make the control of uncertain processes a challenging problem. This paper proposes a discrete-time controller for a class of highly uncertain dynamical systems, whose design is given in two stages: (i) a data-driven model is obtained by means of a fuzzy neural network and a nominal controller for the disturbance-free case is considered; and (ii) a disturbance-observer is formulated to compensate unknown effects. The boundedness of the tracking error signals is analysed and the performance of the proposed scheme is validated throughout experimental results.