Design of an adaptive neuro-observer-based feedback controller for high-speed trains with parametric uncertainties

An adaptive observer-based controller design for the nonlinear model of a high-speed train is demonstrated in this paper. A high-speed train belongs to the class of multivariable and coupled dynamic system with a higher degree of nonlinearities and uncertainties. The radial basis function neural network is used for the approximation of these nonlinearities and uncertainties. Using this approximation, an adaptive neuro-observer is designed for the estimation of the unavailable states of the high-speed train for measurement. Using one-to-one nonlinear mapping, the high-speed train plant and the adaptive neuro-observer are remodelled in a pure feedback form without any constraints on states. On the basis of the adaptive neuro-observer, an adaptive dynamic surface controller is designed for the high-speed train system. The upper bounds of the actual controller gains of the high-speed train need not be known whereas the lower and upper bounds of the virtual controller gain require prior knowledge. The tuning of the design parameters has been done online in the proposed observer/controller. The closed-loop stability and convergence have been analysed through a formal proof based on the Lyapunov approach. The enhanced performance of the high-speed train with the proposed controller is compared with the backstepping control approach and demonstrated using simulation studies.