To address the interference caused by six-degree-of-freedom wave-induced motions during the landing of shipborne unmanned aerial vehicles (UAVs) at sea, this study proposes a double-layered Stewart platform compensation mechanism. A kinematic model of the platform is constructed using a closed-loop vector method, enabling coordinated control in which the lower layer simulates wave disturbances and the upper layer performs dynamic compensation. To mitigate the effects of sensor delay and noise, a hybrid CNN-BiLSTM prediction model is developed. Using ship pitch attitude as an example, the model significantly improves the accuracy of attitude prediction. Experimental results demonstrate that the proposed approach achieves effective wave compensation, offering a viable solution for UAV landings at sea.