Comparison of two-level NMPC and ILC strategies for wet-clutch control

Modeling and control of clutch engagements has been recognized as a challenging problem, due to nonlinear and time-varying dynamics, switching discontinuously between two phases. Furthermore, the optimal references are not known a priori and vary with operating conditions. To address these issues a two-level control scheme is proposed, consisting of a learning algorithm at the high level, updating parameterized references to be tracked at the low level. To simplify the tracking, the controls for both phases are separated. In a first implementation, two (non)linear model predictive controllers (NMPCs) are used sequentially, while in a second implementation these are replaced by two Iterative Learning Controllers (ILCs). The performance and robustness are investigated on a test setup with wet-clutches, and it is shown that both implementations combined with suitable high level algorithms result in good engagements. (C) 2013 Elsevier Ltd. All rights reserved.