基于改進Seq2Seq的船舶軌跡預測模型

打開文本圖片集

中圖分類號：U675.7 文獻標志碼：A

Abstract：Aiming at the problem that the traditional recurrent neural network （RNN）model has slow convergence speed and low accuracy，resulting in a large diference between the predicted trajectory and the real trajectory of a maritime ship，an Seq2Seq（sequence to sequence）model composed of RNNs is constructed.Attention mechanism and convolutional neural network （CNN）are introduced to improve the model，strengthening the ability of extracting data features，accelerating the convergence speed of the model，and improving the trajectory prediction accuracy. The experimental results show that：compared with the traditional RNN model，the mean square error，the root mean square error，and the average absolute error of the Seq2Seq model are reduced by 81.41% ， 12.67% ，and 62.43% ，respectively ; compared with the Seq2Seq model，the mean square error，the root mean square error，and the average absolute error of the improved Seq2Seq model are reduced by 42.87% ，69. 27% ，and 45. 79% ， respectively.

Key words： ship trajectory prediction；Seq2Seq （sequence to sequence）；attntion mechanism;convolutional neural network （CNN）；recurrent neural network （RNN）

0 引言

近年來，隨著經(jīng)濟和對外貿易的發(fā)展，船舶數(shù)量不斷增多，但也帶來了航行效率以及航行安全等問題[1]。（剩余7731字）