ST-Seq2Seq: A Spatio-Temporal Feature-Optimized Seq2Seq Model for Short-Term Vessel Trajectory Prediction

Abstract

Deep learning provides appropriate mechanisms to predict vessel trajectories for safer and efficient shipping, but still existing models are mainly oriented to longer-term prediction trends and do not fully support real time navigation needs. While most recent works have been largely exploiting Automatic Identification System (AIS), the complete semantics of these data haven't so far fully exploited. The research presented in this paper introduced an extended sequence-to-sequence model using AIS data. A Gated Recurrent Unit (GRU) network encodes historical spatio-temporal sequences as a context vector, which not only preserves the sequential relationships among trajectory locations, but also alleviates the gradient descent problem. The GRU network acts as a decoder, outputting target trajectory location sequences. Real AIS data from the Chongqing and Wuhan sections of the Yangzi River were selected as typical experimental areas for evaluation purposes. The proposed ST-Seq2Seq model has been tested against the LSTM-RNN and GRU-RNN baseline models for short term trajectory prediction experiments. A 10-minute historical trajectory sequence was used to predict the trajectory sequence for the next five minutes. Overall, the findings show that LSTM and GRU networks, while applying a recursive method to predict a sequence of continuous trajectory points, when the number of predicted trajectory points increases accuracy decreases. Conversely, the extended sequence-to-sequence model shows satisfactory stability on different ship channels.

FRA-LSTM: A Vessel Trajectory Prediction Method Based on Fusion of the Forward and Reverse Sub-Network

“…However, the literature [5] applied the BLSTM network to retrieve the abnormal value devalued trajectory points, not consider the influence of the backward existing trajectory data on the target recovery trajectory data. We reproduced the vessel trajectory prediction work of BLSTM [5] and Seq2seq [24], and their results proved that they were only effective for short-term prediction, but the accuracy of long-term trajectory prediction was significantly reduced. Hence, how to eliminate the above shortcomings, consider the influence of navigation behavior on trajectory prediction, and improve the accuracy of vessel trajectory prediction, which is the motivation of this paper.…”

“…The main idea of the current vessel trajectory prediction research is to analyze and mine the AIS data. You et al [24] in 2020 introduced an extended sequence-to-sequence model based on AIS data, which implemented a recurring grid unit (GRU) network to encode historical vessel spatiotemporal sequences into context vectors. The model preserved the order relationship between trajectory positions and solved the problem of gradient descent.…”

“…The GRU network was adopted as a decoder to output the target trajectory positioning sequence. However, the literature [24] only considered the sequence relationship between the positions, which did not consider the influence of the vessel's motion behaviors such as acceleration and steering on the trajectory prediction during the traveling process [24]. In 2020, to improve the accuracy of vessel trajectory prediction, Liu et al developed a two-stage data-driven machine learning hull trajectory reconstruction framework [5].…”

AIS-Based Intelligent Vessel Trajectory Prediction Using Bi-LSTM

“…This helps us to improve the accuracy of neural network prediction and to establish different prediction models for different vessel trajectories. We can predict the trajectories more accurately under similar navigation modes [56].…”

Ship Track Prediction Based on DLGWO-SVR

“…is method improves the prediction accuracy to some extent, but it is still not ideal. Prediction methods based on the machine learning model include recurrent neural network [6], backpropagation neural network [7], long and short time memory network [8], and Support Vector Regression (SVR) [9]. With the following characteristics: As data amount increases, gradient explosion or gradient disappearance may occur, and the convergence rate is slow, which leads to the decrease of prediction accuracy and the low efficiency of sample training.…”