On Some Issues Of Automatic Train Control Based On Iterative Learning Control

ABSTRACT:This dissertation introduces iterative learning control (ILC) into automatic train control field and intensively studies several important problems, including dynamical model identification, trajectory tracking control, safety control and station stopping control. Some novel algorithms are proposed, which make full use of the repeatability of the train motion pattern in order that the control performance can be iteratively improved through the repetitive running cycles.The main works and contributions are summarized as following:1. An iterative learning identifier is developed to identify the parameters in the train dynamical model. The identifier has an ability to make the parameters converge to their actual values through repeated identifying trials where the experimental data serve as the desired outputs and the parameters to be identified serve as the control inputs. Both the theoretical analysis and simulation examples demonstrate the validity and effectiveness of the proposed identification method.2. An iterative learning control approach is proposed to address the trajectory tracking problem of a train operation. The ILC-ba33sed controller makes use of previous speed tracking error to modify the current control input (tration force or braking force). Therefore, the controlled train is guranteed to track the desired trajectory (usually from optimal scheduling) without deviation when the running cycle increases to infinity. Finally, the simulation examples verify the effectiveness of proposed algorithms.3. The safety issues under the ILC-based trajectory tracking algorithm are discussed in detail. With theoretical analysis of the speed and displacement tracking errors, the sufficient conditions to prevent overspeed and collision accidents are derived. Finally, the minimum headway problem is investigated.4. Three train station stop control algrithms based on terminal iterative learning control (TILC) are proposed. The initial braking point, the braking force and the combination of them are chosen as the control profile in turn, and the corresponding learning laws and convergence theorems are presented respectively for the three scenarios. In the 3rd scenario, the initial state (initial braking point) and the external input signal (the braking force) are chosen as the control input simultaneously, in order that the convergence speed is effectively increased. This provides a new structure of the terminal iterative learning controller.In these ILC-based train control algorithms, the control tasks are technically extended from the time domain to the iteration domain. Thus, the control performance can be effectively improved along the iteration axis by utilizing the obvious repeatability. This is a main priority compared with other control methods. Meanwhile, some proposed algorithms also make progress in ILC theory.