Abstract: A model reference adaptive control system with nonlinear compensation for robot manipulator is proposed in this paper. The adaptive control is based on Liapunov direct method and guarantees the global statility of the system. Also, the method does not require detailed modelling of manipulator dynamics and thus reduces the amount of realtime computation consideiably. Computer simulations and experiments show that the proposed adaptive control system drives the manipulator to follow the reference model closely in different payload situations. The proposed nonlinear compensation method cancels the inhei-ent nonlinearity of Coulomb friction quite effectively. Comparison shows that the proposed adaptive control is much better than PID control.