Nonlinear generalized predictive control based on online least squares support vector machines

Zhenkai Guo,Xinping Guan
DOI: https://doi.org/10.1007/s11071-014-1733-2
2015-01-01
ACTA AUTOMATICA SINICA
Abstract:For a class of nonlinear discrete systems with unknown parameters, an adaptive direct generalized predictive control method based on online least squares support vector machines (OLS–SVM) is proposed. In the method, the OLS–SVM is used to design the controller directly, and an improved projection algorithm based on the tracking error is introduced to adjust the weights of the OLS–SVM adaptively, so the inverse matrix is avoided in the process of online real-time control. It is proved that the proposed method can make the tracking error converge to a small neighborhood of the origin. Simulation results have shown the correctness and effectiveness of the proposed method.
What problem does this paper attempt to address?