Method for Select Best AIS Data in Prediction Vessel Movements and Route Estimation

Abstract

The prediction of vessel maritime navigation has become an exciting topic in the last years, especially considering economics, commercial exchange, and security. In addition, vessel monitoring requires better systems and techniques that help enterprises and governments to protect their interests. Specifically, the prediction of vessel movements is essential for safety and tracking. However, the applications of prediction techniques have a high cost related to computational efficiency and low resource saving. This article presents a sample method to select historical data on vessel-specific routes to optimize the computational performance of the prediction of vessel positions and route estimation in real-time. These historical navigation data can help to estimate a complete path and perform vessel position predictions through time. This Select Best AIS Data in Prediction Vessel Movements and Route Estimation (PreMovEst) method works in a Vessel Traffic Service database to save computational resources when predictions or route estimations are executed. This article discusses AIS data and the artificial neural network. This work aims to present a prediction model that correctly predicts the physical movement in the route. It supports path planning for the Vessel Traffic Service. After testing the method, the results obtained for route estimation have a precision of 76.15%, and those for vessel position predictions through time have an accuracy of 81.043%.

Special Issue on Interdisciplinary Artificial Intelligence: Methods and Applications of Nature-Inspired Computing

“…Other works of applications using metaheuristic optimization methods are included in the Special Issue. In the work of Bautista-Sánchez et al [6], the authors present a method called prediction vessel movements and route estimation, for selecting the historical data required to compute an optimal path planning of route vessels and an estimation of vessels' positions.…”

Long-Term Trajectory Prediction for Oil Tankers via Grid-Based Clustering

“…Existing short-term prediction algorithms may be used to produce long-term predictions through recursions; however, the accuracies are expected to decrease as the number of prediction steps increases [6]. One relevant work that tackles the problem of long-term trajectory can be found in [22], where DBSCAN was used to cluster historical trajectories, and the trajectory predictions are produced by pretrained deep learning models. This work was tested using the publicly available AIS data from MarineCadastre in Zones 15 and 16.…”

“…This work was tested using the publicly available AIS data from MarineCadastre in Zones 15 and 16. While [22] also adopted DBSCAN, the overall methodology is completely different from our proposed approach. For instance, in [22], DBSCAN was used to cluster trajectories based on trajectory statistics such as the trimmed mean of the longitude (latitude).…”

“…While [22] also adopted DBSCAN, the overall methodology is completely different from our proposed approach. For instance, in [22], DBSCAN was used to cluster trajectories based on trajectory statistics such as the trimmed mean of the longitude (latitude). In contrast, our work adopts DBSCAN to cluster the key points from many historical trajectories, where key points sharing similar characteristics are grouped into the same cluster.…”

Unsupervised vessel trajectory reconstruction

“…In this situation, there is no training set for applying supervised methods to identify the vessel and predict trajectories, and hence unsupervised methods are required. Although the existing unsupervised clustering methods can be used for predicting trajectories of vessels, they may not be able to provide desired prediction accuracy [4]. We propose an unsupervised trajectory reconstruction method can be used for space debris path prediction since space debris typically lack known labels for model training [5], and analyze and investigate three AIS datasets provided by NSF's ATD program and collected from the 1st of June to the 31st of July, 2019 (see Table 1).…”