Intelligent Data-Driven Vessel Trajectory Prediction in Marine Transportation Cyber-Physical System

Abstract

With the rapid developments of artificial intelligence (AI) and Internet of Things (IoT), a large number of autonomous and unmanned surface vessels will appear in the emerging marine transportation cyber-physical system (mTCPS). However, the tremendous growth of moving vessels could lead to unsatisfactory maritime safety and efficacy. To promote smart traffic services in IoT-enabled mTCPS, it is necessary to accurately and robustly predict the locations of moving vessels (i.e., vessel trajectories). The prediction results are beneficial for avoiding vessel collision, detecting abnormal behaviors, and improving maritime surveillance, etc. Motivated by the strong learning capacity of deep learning, this work proposes an intelligent data-driven vessel trajectory prediction framework based on the famous long short term memory (LSTM). To achieve accurate trajectory prediction, the vessel traffic conflict situation modeling, generated using the dynamic positioning data and social force concept, is embedded into the original LSTM network. Furthermore, a mixed loss function is reconstructed to make our trajectory prediction framework more reliable and robust under different experimental conditions. Both quantitative and qualitative experiments on realistic vessel trajectories have been implemented to demonstrate that our method could achieve superior trajectory prediction in terms of accuracy and robustness.