Ensemble-trained neural networks predict ship maneuvering motion and produce uncertainty estimates that grow when the vessel leaves the training data distribution.
Quantitative Evaluation of Full-Scale Ship Maneuvering Characteristics During Berthing and Unberthing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
Leveraging empirical data is crucial in the development of accurate and reliable virtual models for the advancement of autonomous ship technologies and the optimization of port operations. This study presents an in-depth analysis of ship berthing and unberthing maneuvering characteristics by utilizing a comprehensive dataset encompassing the operation of a full-scale ship in diverse infrastructural and environmental conditions. Various statistical techniques and time-series analysis were employed to process and interpret the operational data. A systematic analysis was conducted on key performance variables, including approach speed, drift angles, turning motions, distance from obstacles, and actuator utilization. The results demonstrate significant discrepancies between the empirical data and the established maneuvering characteristics. These findings have the potential to significantly enhance the accuracy and reliability of conventional maneuvering models, such as the Mathematical Modeling Group (MMG) model, and improve the conditions used in captive model tests for the identification of maneuvering model parameters. Furthermore, these findings could inform the development of more robust autonomous berthing and unberthing algorithms and digital twins.
fields
eess.SY 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Probabilistic Prediction of Ship Maneuvering Motion using Ensemble Learning with Feedforward Neural Networks
Ensemble-trained neural networks predict ship maneuvering motion and produce uncertainty estimates that grow when the vessel leaves the training data distribution.