Data-driven Terminal Iterative Learning Control With Variant Uncertainties

This paper mainly focuses on the exploration of some new data driven terminal iterative learning control(TILC) methods and its uncertain problems for discrete-time systems. The main contributions are summarized as follows:First, a high order data-driven optimal terminal iterative learning control algorithm(H-DDOTILC) is proposed for a class of single-input-single-output(SISO) nonlinear systems, where the control input is required to be constant in the same iteration. The proposed H-DDOTILC can utilize more I/O data of previous operations of a repetitive process to update the control input signals to improve the convergence performance along the iteration direction. Further, the result is extended to multiple-input-multiple-output(MIMO) nonlinear systems with time-varying input signals. Both theoretical analysis and simulation results have confirmed the effectiveness of the proposed methods.Second, a new RBF based terminal iterative learning control is proposed with a neural network initial state learning mechanism to deal with the non-repetitive problems of iteration-varying initial states and desired terminal points, which are always exist in practical applications. Even if both the initial states and desired terminal points vary with iterations, the proposed method can still guarantee that the terminal tracking error converges to a predefined bound. It is worth pointing out that the proposed controller does not include any model-information of the controlled process explicitly but uses the system I/O data only. Therefore, the proposed method is also applicable for nonlinear system bypassing modeling steps although the design and analysis is based on a linear system.Third, when the initial states are not measurable, a new data-driven optimal terminal iterative learning control algorithm with initial value dynamic compensation is proposed for a class of nonlinear and non-affine systems. Although the iteration-varying uncertainties exist in initial states and exogenous disturbances, the proposed method can guarantee its robustness and adaptations. So, the proposed approach relaxes the strictly repetitive conditions required in traditional ILC. Robustness analysis and simulation study have verified the effectiveness of the proposed method.