Approximate feedback linearization of nonlinear systems: a data-based direct design approach

The design of feedback-linearizing controllers for discrete-time nonlinear plants, using input/output/state measurements only, is typically addressed via indirect design. In this paper we propose the use of a new technique, based on a virtual reference direct design approach. Such a technique reduces the control design problem into a standard nonlinear approximation problem, without calling for the preliminary identification of the model of the plant. As compared with the existing data-based methods, the new approach requires less computational effort, while taking full advantage of the nonlinear approximation software tools already available.