Predicting a vessel Estimated Time of Arrival (ETA) while traveling on waterways can be a challenging endeavor in an Intelligent Transportation System (ITS). Moreover, predicting accurate ETAs of a vessel on inland waterways plays a vital role in the inland ports to enhance efficiency and automation, logistics and supply chain process, reduce costs, and improve port users satisfaction. In this paper, using Gradient Boosting Decision Trees (GBDT), Multi-Layer Perceptron Neural Network (MLP), and Gated Recurrent Unit Neural Network (GRU) algorithms trained on past inland waterway AIS data to predict ETAs. These algorithms can deal a large number of AIS data generated by individual vessel trips in inland natural and artificial waterways. In this work, the ability to predict accurate ETA using GBDT, MLP, and GRU are compared and evaluated using historical test data of both inland natural and artificial waterways. GRU provides the best prediction accuracy, with the Root Mean Square Error (RMSE) of 8.50 minutes and Mean Absolute Error (MAE) of 5.92 minutes. Furthermore, it is shown that the GRU model performs better in artificial waterways than natural waterways. The model shows the prediction error decreases as the distance of the waterway increases.
Towards Data-Driven GRU based ETA Prediction Approach for Vessels on both Inland Natural and Artificial Waterways
2021-09-19
2107295 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Scheduling Vessels in Natural Inland Waterways
Springer Verlag | 2024
|Fast Vessels on Inland Waterways
British Library Conference Proceedings | 1999
|Feature 5 - Inland vessels - Diverse requirements and vessels for inland waterways
Online Contents | 2009
Lubrication oil usage on inland waterways vessels
Kraftfahrwesen | 1985
|Lubricating Oil Usage on Inland Waterways Vessels
SAE Technical Papers | 1985
|