Abstract: The underwater snake robot is a new type of bio-inspired unmanned underwater vehicle. Because snake robots are made of rigid materials, the utilization of articulated joints in a snake robot reduces the degrees of freedom, increases the probability of collision, and ultimately leads to damage of these rigid materials in the complex underwater environment. To resolve such an inevitable issue, this work proposes an innovative approach to tailor the structural composition of the robot, which consists of four different modules: the head module, control module, thruster module, and soft joint module. Thus, we formulate a rigid–soft unified mathematical model by jointly modeling the two types of modules with large differences in their physical characteristics. The snake robot is also known as a typical redundant drive robot, which can use redundant degrees of freedom to complete the multitask control. This establishes a task-priority control method based on the rigid–soft unified model. Therefore, the effectiveness of the controller can be verified by simulating the trajectory tracking control of both ends of the robot in different situations, with the end position and attitude control as the primary task and the base position control as the secondary task. This simulation scenario primarily covers the following three situations: The first situation is that the base remains fixed, and the end tracks a predetermined trajectory. In this case, the snake robot can imitate an industrial mechanical arm to perform terminal tasks in water, such as grasping. The second case is that while the end is tracking a predetermined trajectory, the base also moves according to the predetermined trajectory. The simulation of this situation is indicative of the movement process (e.g., turning) of the snake robot, enabling obstacle avoidance in spacious waters. In the third case, the attitude control of the robot in a fixed position has been verified through simulation studies. We anticipate that both the base and end of the robot will remain in the same position, and the attitude of the end will move according to a predetermined angle. This situation is to simulate the snake robot for capturing more visual information in a narrow underwater environment. The results revealed that when the base is fixed and the end moves, the average tracking error of the end is 1.41% in the X direction and 1.02% in the Y direction. Similarly, when both ends move simultaneously, the average tracking error of the end is 0.53% in the X direction and 1.64% in the Y direction. Moreover, when the terminal attitude is controlled, the average error of pitch angle is 0.38%, and the average error of yaw angle is 0.14%, which verifies the effectiveness of the controller. Such a beneficial performance can considerably reduce the damage of snake robots in complex underwater treatment by means of improving their swimming efficiency and flexibility.