Foldable parallel mechanisms are suitable to compensate for the motion of vessels due to their compact structure, multidimensional motion, and large carrying capacity. However, the vessel sways violently under harsh sea conditions, challenging the rational utilization of the mechanism workspace. This work proposes a novel motion planning algorithm based on motion prediction and the B-spline for a three-degrees-of-freedom foldable parallel compensation platform. First, for harsh sea conditions, the motion law of a guardian vessel is studied, and the encoder-decoder model is used to predict the roll, pitch, and heave motions of the vessel. Second, we propose a multidimensional motion compensation strategy under the limits of the mechanism dimensions. Then, a backward kinematics model is developed. Its workspace under the structural and actuator limitations is proposed. Finally, based on the theory of the nonuniform rational B-spline curve and target motion prediction, a dynamic motion planning algorithm is proposed, and experiments are conducted. The motion planning algorithm can provide smooth motion commands under the mechanism workspace, velocity, acceleration constraints. It can be realized using asynchronous calculation, which meets real-time motion control requirements.