Neural networks possess a nonlinear approximationproperty, so they can be employed in direct adaptivecontrol for flexible manipulators. In this thesis, arigid model-based neural network control is proposed.The structure of controller is composed of two parts.The first part is designed from the rigid model and thesecond part, which comprises a neural network, is usedto control the flexible part of the flexiblemanipulator. A major advantage of the neural networkadaptive controller design over previous ones is thatit does not need the linearity-in-the-parametersassumption. Moreover, model details that are hard toobtain exactly, especially in case of flexiblemanipulator, are not necessary. In addition, since theparameter adaptation law is obtained from the Lyapunovapproach, the neural networks learn on-line in realtime with no off-line training needed. Both thecomputer simulation and the experimental result ofsingle-link flexible manipulator control show that theproposed controller can be used satisfactorily.