Model predictive control (MPC) is an optimization-based approach and decides a control input for the system output to track a reference trajectory using an optimal computation. The reference trajectory is an ideal trajectory for the system output to converge on a desired value. In this paper, a tracking controller for a two-link planar manipulator on the horizontal space via nonlinear model predictive control (NMPC) is proposed. In order to guarantee a desired tracking performance, a time-variable and time-coefficient of the reference trajectory is used in the proposed controller instead of using a time-constant. The time-coefficient is tuned based on a control error between a controlled variable and a desired value. In order to show the effectiveness of the proposed NMPC controller, numerical simulations comparing the proposed method with the previous method are performed.