Constrained Data-Driven Iterative Learning Control
Ronghu Chi,Yu Hui,Zhongsheng Hou
DOI: https://doi.org/10.1007/978-981-19-5950-9_10
2022-01-01
Abstract:In real process industries, a soft limiter is often added directly into the controller to prevent over large control input signals (Doyle, 1999). However, such a limiter makes the control systems highly nonlinear and may introduce instability. It is also worth pointing out that the input saturation will bring more serious influence on ILC systems because ILC is virtually an integral action along the iteration direction (Xu et al. 2004). Recently, several ILC schemes with input saturation have been proposed under the framework of contraction mapping (Sun and Wang 2001; Driessen and Sadegh 2002; Chu and Owens 2010; Huang et al. 2013). However, the contraction mapping based ILC may cause some poor transient performance, even divergence at some time instants. Therefore, an adaptive ILC with input saturation has been proposed in Xu et al. (2004) by introducing a composite energy function. In Tan et al. (2011), a ‘reference governor’ based ILC is proposed for input constraint systems where the reference signal can be re-designed such that the tracking objectives realizable. The authors in Tan et al. (2012) propose a unified design framework of ILC to deal with nonlinear input uncertainties including input saturations. Besides the input saturation, the output constraints are also of a great concern from a practical viewpoint. For example, the train operation system is constrained not only by the input force because of its mechanical features, but also by its operation speed because of its over speed limits. The latter seems more important to ensure operation performance and safety. Generally, the position and/or speed constraints are required in most motion systems. In the expository overview of ILC (Moore, 1999), optimal ILC (Amann et al. 1996a, b, c; Lee et al. 2000) is revisited and classified as a connection between ILC and optimal control. By using a predesigned objective function, various practical issues such as constraints on the system output and control input, disturbances, measurement noises, and model errors can be considered in a rigorous and systematic manner. Recently, some model predictive control (MPC)-based ILC schemes are also proposed in the literatures (Mishra et al. 2010; Amann et al. 1998; Lee et al. 1999; Wang et al. 2013; Chin et al. 2004) by incorporating the input and output constraints into the optimization problem.