Abstract: A control system design method based on dynamic interpolation strategy is proposed to improve the motion control performance of rehabilitation robot. Firstly, the impaired-limb movement features are extracted with sliding standard deviation for a certain data samples, and the real-time physical state of impaired limb is acquired by fuzzy reasoning. Secondly, the appropriate interpolation method for motion control is recommended using dynamic interpolation decision-making mechanism. Finally, the position-based impedance control is adopted to carry out training exercise. The proposed method avoids the limitations of the conventional single-interpolation methods, and effectively integrates the characteristics of different interpolation methods. Experimental results based on WAM^TM rehabilitation robot show that the designed control system with proposed method demonstrates better motion control performances under disturbances.