Adaptive control of spacecraft and large structures using Radial Basis Function (RBF) based Artificial Neural Networks (ANN) is investigated. The centers of approximation of the RBFs are allowed to be dynamic in order to provide persistent excitation and a small window of approximation. Both state and time based RBFs are investigated for their ability to identify unmodeled and persistent effects. Integral feedback of the attitude seems necessary, especially when the initial estimates are poor, to eliminate steady state errors for pointing applications. Examples of motion-to-rest as well as tracking maneuvers and vibration suppression are presented.