Improved model‐free adaptive predictive control based on recursive least‐squares estimation algorithm
Hoang Anh Pham, Dirk Söffker
Abstract
Abstract Data‐driven control has received increasing attention by researchers in recent years because no system modeling procedure is required. By combination of data‐driven and model predictive control, this paper discusses an improved model‐free adaptive predictive control approach with application to vibration reduction of an elastic crane. For system linearization, instead of using traditional compact‐form dynamic linearization, this contribution considers the partial‐form dynamic linearization (PFDL) technique in case of multivariable systems. A linearized output predictive model of the unknown system is constructed locally. To estimate and predict the time‐varying parameter matrices, namely, pseudo‐Jacobian matrix (PJM), recursive least‐squares algorithms are utilized for online estimation improvement. As a result, an improved PFDL‐based model‐free controller is designed and applied firstly to the ship‐mounted boom crane representing a typical flexible system. Simulation results indicate that significant reduction of the elastic boom and payload oscillations are achieved, and better control performance can be observed in comparison with other traditional methods.