In this paper, a model predictive control algorithm is developed for the regulation problem of a biaxial feed drive system. The orthogonal error component in the moving frame is considered as an approximation of the real contour error. Then, the control policy is derived from the worst-case optimization of a quadratic cost function, which penalizes transformed errors, velocity errors and control variables in each sampling time over a finite horizon. In addition, the constraint is satisfied to ensure the convergence against uncertain but bounded disturbances. The good performance of the proposed control algorithm is verified via computer simulations with predefined trajectories. Furthermore, the result shows the improvement of the tracking accuracy by comparing with the unconstrained predictive control methods.