Neural network augmented anti-skid controller for transport aircraft

This paper describes the design and simulation testing of an anti-skid brake controller for brake-by-wire systems of transport aircraft. The proposed design is based on on-line identification of the friction properties of the runway-tire interface, achieved by combining a multilayer perceptron neural network with fast parameter estimation. We employ recent results from robust control theory to guarantee stability of the whole system despite large uncertainties on the hydraulic components and the brake torque function. In contrast to most commercially available anti-skid controllers, the control signal is smooth, does not result in pressure build-dump cycles and keeps the wheel slip below the tread adhesion limit, thereby potentially reducing tire wear, in addition to improving stopping efficiency.