Design of Model-Free Adaptive Predictive Controllers with Low Online Computational Load for a Class of Discrete-Time Nonlinear Multiple-Input Multiple-Output Systems

Abstract

This paper proposes a novel model-free adaptive predictive control approach with low online computational load for of a class of unknown discretetime nonlinear multiple-input multiple-output systems. First, historical measurable input/output data is utilized to online construct equivalent explicit data models for control design, making the controllers model-free. Second, input structuralization where system inputs are expressed by linear combination of a few basis functions is integrated into predictive controller design for reducing online computational load. Third, a novel Kalman Filter-based regression factor computation approach is developed to predict data model parameters for realizing predictive function of controllers such that closed-loop system robustness can be enhanced when sudden changes from prescribed references are encountered. Last, numerical simulations validate that, by using much less online computational load, the proposed approach can achieve equivalent control performance compared with the existing approaches.