This paper deals with the problem of decentralized stabilization for a class of unknown large-scale dynamic systems formed of interconnected subsystems under constant disturbances. A decentralized adaptive control scheme is proposed based on the Lyapunov direct method and a condition of asymptotic stability is derived. Then the decentralized adaptive control scheme is applied for a class of robot manipulators to track the desired trajectory as closely as possible despite a wide range of manipulator motions and parameter uncertainties of links and payload.