This paper presents a dynamic velocity planning method for tracking error prediction and control based on a theoretical model of a servo control system. A balance between machining efficiency and machining accuracy is achieved by the proposed approach compared with the traditional method of reducing kinematic constraints and sacrificing machining efficiency to improve machining accuracy. Firstly, this study introduces a timeoptimised algorithm based on velocity, acceleration and jerk and uses the pseudo-jump method to linearise nonlinear problems. Secondly, based on the three-loop model of PID servo control system commonly used in industry, a new tracking error prediction method based on kinematic parameters is proposed. A new feed rate planning control algorithm based on dynamic characteristics is proposed to realise the control of tracking error, considering the dynamic performance of the axis and based on the prediction results. Furthermore, because the speed control algorithm based on dynamic characteristics is not a linear constraint, this study proposes a method combining convex relaxation and pseudo jerk to linearise the nonlinear problem, effectively obtaining a near-optimal feed rate profile. Finally, on the basis of these research contributions, the experiments are compared with the traditional kinematics method. Results show that the new method has less tracking error and higher processing efficiency.