Abstract. Input shaping technique can be used to suppress residual vibration, occurring from moving rapidly a flexible system from one point to another point. An input shaping filter produces a shaped input signal that avoids exciting the flexible modes of the flexible system. The technique requires accurate knowledge of mode parameters. When the plant model is not accurate, performance of the input shaper degrades. Several robust input shapers were proposed to handle this inaccuracy at the expense of longer move time. The purpose of this paper is, for the first time, to present an application of an intelligent backstepping system to matching of the resulting closed-loop system with a reference model. The input shaper can then be designed from the mode parameters of the reference model. Because the reference model is accurate even when the plant model is not, the input shaper needs not be robust, resulting in shorter move time. The intelligent backstepping system consists of a three-layer neural network, a variable structure controller, and a backstepping controller. The neural network is used as a black-box model in case when the plant model is unknown, making the proposed system model-independent. The adaptive property of the neural network also makes the proposed system suitable for nonlinear, time-varying, or configuration-dependent systems. The variable structure controller handles the uncertainty arisen in the system. The backstepping controller, through its virtual controls, provides a means for the control authority to reach the unmatched uncertainty in the system. This study contains simulation and experimental results on a flexible-joint robot manipulator. The results showed that this proposed intelligent input shaping system outperformed previously proposed robust input shapers in terms of allowable uncertainty amount and move time. The proposed system is also relatively easy to apply because it does not require the plant model.