Global composite learning velocity tracking control for heavy haul trains

In this paper, the velocity tracking control problem of heavy haul trains (HHTs) is focused on multibody and control-oriented dynamic models. A composite learning algorithm is developed by combining a neural network with a high-order disturbance observer to approximate unknown nonlinearities and compounded disturbances collaboratively, where a prediction error is introduced to assess and improve the approximation accuracy. Since the neural network approximation ability holds only over a compact set, neural network-based control schemes can only ensure semi-globally uniform ultimate boundedness. Thus, a global tracking control scheme for HHTs is proposed that can switch between the composite learning controller and an additional robust controller via a switching mechanism. Finally, the globally uniform ultimate boundedness of closed-loop system signals is proven through Lyapunov theory. Simulation experiments are carried out based on an HXD1-type HHT running on the Da-Qin Line in China, and the results demonstrate the effectiveness of the proposed models and control technique.