Aerodynamic parameters of hypersonic flight vehicle are extremely sensitive to flight condition for its high flight velocity. These usually make the model parameters vary in large range and normal robust control method invalid. This paper evaluates the model uncertainties by traversing the variation values of the flight states. Neural network control is introduced to eliminate the large variations of the model parameters and neural network-based robust controller is designed to realise the tracking control of hypersonic flight vehicle. The simulation results demonstrate the validity of the proposed approach.