On Some Issues Of Improved Iterative Learning Control And Its Application To Train Operation Control

ABSTRACT:This thesis focuses on the improved iterative learning control (ILC)methods and their applicationson automatic train control. The main research contents and innovations are summarized as follows:1. Through analyzing, it is found that existingfeedback control based automatic train control methods are lack of ability to utilize the repetition of train operation to improve control performance.Therefore, a feedback-feedforward ILC control method is designed for train trajectory tracking problem,whose convergence is analyzedfor the systems both with and without input constraint and measurement noises. The proposed approach could benefit from the performance improvement afforded by the learned feedforward as well as robustness benefits afforded by the feedback element;2. The computational complexity and therequired large memory make existing2-norm optimal ILC (NOILC) not practicable. Thus a computationally efficient NOILC (E-NOILC) is developed fordiscrete-time linear time-varying systems. Its propertiesare analyzed in theory, which are convergence, robustness, convergence rate, control performance, and computational complexity. The effectiveness is verified by experiments on motor motion control testbed. The lifted technique is not applied in the controller design, which reduces the computational complexity of existing NOILC and widen the range of applications;3. Considering the cooperative control for multi-agent group, the computationally efficient norm optimal iterative learning cooperative control (coE-NOILC) is studied for a group of discrete-time linear time-varying systems.The control method combines the individual ILC and cooperative ILC into an optimal ILC framework. In addition, properties of the controllerare provided, includingconvergence, robustness, convergence rate, and control performance. Compared with existing coordinated ILC, the transient behavior along iteration axis is improved by introducing optimization;4. To improve the control performance of train operation, abovementioned E-NOILC and coE-NOILC methods are extended to input-affine nonlinear systems, and applied to train trajectory tracking problems;5. Considering the potential security problemsin train operation, the ILCalgorithmwith overspeed protection is studied by adding a state feedback term, and the ILC algorithm with collision protection is researched by adding an output cooperation term.These two control methods consider safety problems, which enhances the safety of train operation system.