Abstract: In order to assist the surgeons to complete the fine operation of retinal microsurgery, a method of generating spatial motion constraints for surgical robots, named virtual fixture (VF), is proposed to filter the trembling, and improve the accuracy and stability. Firstly, the surgical environment constraints and the task constraints are introduced, and 6 VF primitives needed for retinal microsurgery are constructed by using the weighted, linearized, multi-objective constraint conditions, and taking the users' input as the objective function. On this basis, taking the generation of the remote-center-of-motion virtual fixture (RCM VF) as an example, the implementation method of complex constrained motion is deduced by the combination of the constrained movement primitives. The simulation results of each constrained motion primitive algorithm and the complex constrained motion algorithm show that, the surgical instrument can realize the specific constrained motions according to the definition of the VF. Finally, based on the introduction of constrained motion primitives in each surgical procedure, experiments are performed on table tennis and the in vitro porcine eye, which prove that the retinal surgical robot can perform difficult surgical operations under the guidance of the VFs. Therefore, the rationality and effectiveness of the proposed algorithm are verified.