Real-Time Ship Motion Forecasting Using Deep Learning

Abstract

It is still challenging to continuously observe the marine ship motion in a harsh environmental condition. Combined science and technology assistance makes human maritime activities undergo a revolution with increasing artificial intelligence aboard ships. For instance, it is now possible to embed sensor processing to automatize tasks that humans could perform at a time. As a result, piloting ships among sea waves is a particular interest. In this paper, we address the prediction based ship's motion using deep learning. To create a 3D image, we use a computer graphic software named Blender. It is possible to simulate ships floating at sea level, recording sea surface images and the vessels' motion parameters, considering their pitch and roll. We employ different kind of neural network models for training include convolutional neural network (CNN), the combination of convolutional neural network with long short term memory (LSTM), and the combination of convolutional neural network with a gated recurrent unit (GRU) network. After training, we analyze different models' performance and discuss how different time gap affects models' performances.