This paper presents an effective approach for multi-task control of the underwater biomimetic vehicle-manipulator system. The main idea of this approach lies in organizing and combining the tasks by priorities, while decomposes the coupled relations among the tasks by null-space mapping consecutively avoiding interaction effects. A direct kinematic model is firstly built to describe the motions and states of the end grasper. Then, the robust closed-loop iterative task-priority algorithm is designed. Here, the kinematic singularity and algorithm singularity are modified by the effective methods to ensure the robustness of the algorithm. Eventually, this algorithm is applied to solve the multi-task control problem of the underwater biomimetic vehicle-manipulator system in the grasping mission. Three noteworthy characteristics are discussed dialectically, and the obtained results show the effectiveness of this algorithm.