Abstract: In this paper,a general learning algorithm for dynamic control of robot manipulator is proposed.In this algorithm,information learned previously can be applied to new control objectives as long as similar regions ofthe system state space are involved.The control scheme requires no prior knowledge of the robot dynamics sothat it is easy to be used to solve a particular control problem or to be modified to accommodate changes inthe physical system.The control scheme is computationally efficient and well suited to fixed-point implementation.The effectiveness of the learning algorithm is verified by the computer simulation experiment of repetitive movements of two fore arms of 4 DOFs direct drive robot manipulator.