针对船舶六自由度运动摇摆台正解问题，本文提出基于注意力机制神经网络的解算方法，利用平台的运动学反解模型构建腿长与位姿的数据集，训练注意力模型以实现高效求解。具体过程包括：首先通过运动学反解的数学模型获取支腿长度与平台位姿的对应关系；其次，利用数据集训练基于注意力机制的神经网络模型；最后，进行仿真验证算法的有效性。仿真结果表明，相较于同样规模的BP神经网络方法，本文提出的方法在训练迭代次数上减少44%，误差降低50%，求解时间缩短53%，求解稳定性增强。

This paper proposes using attention mechanism to replace traditional solving methods for the forward solution problem of ship six degree of freedom motion rocking platform. Construct a dataset of leg length and pose using the kinematic inverse solution model of the platform, and train an attention model to achieve efficient solution. The specific process includes: first, obtaining the corresponding relationship between the length of the support legs and the platform pose through the mathematical model of inverse kinematics; Secondly, training a neural network model based on attention mechanism using the dataset; Finally, conduct simulation experiments to verify the effectiveness of the model. The simulation results show that compared to BP neural network methods of the same scale, the proposed method reduces training iterations by 44%, error by 50%, solution time by 53%, and solution stability is enhanced.