Iterative Learning Control And Its Application In Automatic Train Operation

One major function of automatic train operation is to regulate train speed and guarantee that the train will operate safely with the time table. In this paper, mechanical characteristics of train operation are analyzed and train's dynamic model is presented. Based on iterative learning control theory, four learning laws are proposed, D-type, PD-type, enhanced D-type, and non-parametric adaptive learning law. Based on the learning laws, automatic train operation controllers are constructed. By applying the train's dynamic model, the process of train speed tracking is simulated and the convergence is verified by simulation results. The differences among the four controllers in control capacities are also discussed by analyzing the tracking error and convergence rates.The main contents are as follows:a) Aiming at a class of general linear and nonlinear systems, the techniques of constructing iterative learning controllers are studied. A D-type learning law and the astrictive conditions are proposed. It is shown that after several iterations tracking errors of the system with the same initial conditions converge to zero over a finite interval. Learning laws and astrictive conditions of PD-type, high-order type, and filter-based type iterative learning control are also proposed.b) Based on the basic structure of automatic train operation, a D-type and a PD-type learning law are presented and controllers are established to regulate train speed in the proposed dynamic model. Simulation results indicate that train's speed trajectory converges to the desired speed profile after several iterations and achieves high tracking accuracy. Compared with D-type iterative learning controller, PD-type converges more quickly.c) To reduce the speed tracking error in initial iterations of D-type and PD-type iterative learning controller, an enhanced D-type controller is established. In the new controller, a negative feedback loop is introduced during the initial iteration to optimize the system input. Simulation results show an appealing decrease in the initial speed tracking error and an improvement in control capacity and convergence rates.d) A non-parametric adaptive iterative learning controller is also constructed for automatic train operation to regulate train's speed. From the simulation results, it can be concluded that non-parametric adaptive iterative learning controller achieves higher convergence rates and better control capacities compared with D-type or PD-type iterative learning controller. More importantly, it is much easier to design the learning algorithm for an effective control.