Abstract:On-site estimation of sea state parameters is crucial for ship navigation systems' accuracy, stability, and efficiency. Extensive research has been conducted on model-based estimating methods utilizing only ship motion responses. Model-free approaches based on machine learning (ML) have recently gained popularity, and estimation from time-series of ship motion responses using deep learning (DL) methods has given promising results. Accordingly, in this study, we apply the novel, attention-based neural network (AT-NN) for estimating sea state parameters (wave height, zero-crossing period, and relative wave direction) from raw time-series data of ship pitch, heave, and roll motions. Despite using reduced input data, it has been successfully demonstrated that the proposed approaches by modified state-of-the-art techniques (based on convolutional neural networks (CNN) for regression, multivariate long short-term memory CNN, and sliding puzzle neural network) reduced estimation MSE by 23% and MAE by 16% compared to the original methods. Furthermore, the proposed technique based on AT-NN outperformed all tested methods (original and enhanced), reducing estimation MSE by up to 94% and MAE by up to 70%. Finally, we also proposed a novel approach for interpreting the uncertainty estimation of neural network outputs based on the Monte-Carlo dropout method to enhance the model's trustworthiness.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **Real - time estimation of sea state parameters (such as wave height, zero - crossing period and relative wave direction) through ship motion responses**. Specifically, the author aims to develop an attention - based neural network (AT - NN) to accurately estimate sea state parameters from time - series data of ship motions such as pitch, heave and roll. ### Problem Background 1. **Importance of Sea State Parameters** - Real - time estimation of sea state parameters is crucial for the accuracy, stability and efficiency of ship navigation systems. - Traditional sea state data collection methods (such as moored buoys, navigation radars or satellite data) have problems such as availability, reliability, convenience and maintenance costs. 2. **Limitations of Existing Methods** - Statistical methods can provide accurate wave height estimates, but are poor at estimating wave period and wave direction. - Expensive wave radars can accurately estimate wave period and wave direction, but the wave height estimate is sometimes inaccurate. - Traditional methods based on RAO may be inaccurate or incomplete in some cases. 3. **Advantages of Model - Free Methods** - Using machine learning (ML) and deep learning (DL) methods can estimate sea state parameters from time - series data of ship motions. These methods do not require complex physical models and have higher flexibility and adaptability. ### Core Contributions of the Paper 1. **Proposing a New Attention - based Neural Network (AT - NN)** - This model can directly estimate sea state parameters from the original time - series data, reducing the complexity of the input data. - Compared with existing methods, AT - NN significantly reduces the mean squared error (MSE) and the mean absolute error (MAE) by 94% and 70% respectively. 2. **Improving Existing Deep Learning Models** - By improving the convolutional neural network (CNN), the multivariate long - short - term memory CNN (MLSTM - CNN) and the sliding puzzle neural network (SP - NN), the performance is further enhanced. 3. **Proposing a New Method for Uncertainty Estimation** - Based on the Monte Carlo dropout method, the credibility and interpretability of the model are enhanced. ### Summary The main objective of this paper is to develop an efficient and accurate attention - based neural network for real - time estimation of sea state parameters from ship motion responses, thereby improving the safety and efficiency of ship navigation. By introducing new deep learning models and improving existing models, the author shows significant performance improvements and provides new directions for future research.

Estimation of Sea State Parameters from Ship Motion Responses Using Attention-based Neural Networks

A Novel Densely Connected Convolutional Neural Network for Sea-State Estimation Using Ship Motion Data

A Deep Learning Model for Ship Trajectory Prediction Using Automatic Identification System (AIS) Data.

A Computer Vision Approach to Estimate the Localized Sea State

Response component analysis for sea state estimation using artificial neural networks and vessel response spectral data

An Uncertainty-Aware Hybrid Approach for Sea State Estimation Using Ship Motion Responses

A data driven method for multi-step prediction of ship roll motion in high sea states

State Estimation of an Underwater Markov Chain Maneuvering Target Using Intelligent Computing

A Novel Class-Imbalanced Ship Motion Data-Based Cross-Scale Model for Sea State Estimation

A Novel Deep Learning Model for Sea State Classification Using Visual-Range Sea Images

Data-Driven system identification of 6-DoF ship motion in waves with neural networks

Prediction of wave-induced ship motions based on integrated neural network system and spatiotemporal wave-field data

SAFENESS: A Semi-Supervised Transfer Learning Approach for Sea State Estimation Using Ship Motion Data

Data-Driven Modeling for Transferable Sea State Estimation Between Marine Systems

Towards Improved Prediction of Ship Performance: A Comparative Analysis on In-service Ship Monitoring Data for Modeling the Speed-Power Relation

Nonparametric modeling of ship maneuvering motions in calm water and regular waves based on R-LSTM hybrid method

Ship Anomalous Behavior Detection Using Clustering and Deep Recurrent Neural Network

Ship performance monitoring using machine-learning

Nonparametric modeling of ship maneuvering motion based on self-designed fully connected neural network

Deep learning innovations in South Korean maritime navigation: Enhancing vessel trajectories prediction with AIS data

Probabilistic Maritime Trajectory Prediction in Complex Scenarios Using Deep Learning