Abstract:Machine learning (ML) models often require large volumes of data to learn a given task. However, access and existence of training data can be difficult to acquire due to privacy laws and availability. A solution is to generate synthetic data that represents the real data. In the maritime environment, the ability to generate realistic vessel positional data is important for the development of ML models in ocean areas with scarce amounts of data, such as the Arctic, or for generating an abundance of anomalous or unique events needed for training detection models. This research explores the use of conditional generative adversarial networks (CGAN) to generate vessel displacement tracks over a 24-hour period in a constraint-free environment. The model is trained using Automatic Identification System (AIS) data that contains vessel tracking information. The results show that the CGAN is able to generate vessel displacement tracks for two different vessel types, cargo ships and pleasure crafts, for three months of the year (May, July, and September). To evaluate the usability of the generated data and robustness of the CGAN model, three ML vessel classification models using displacement track data are developed using generated data and tested with real data.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: in the marine environment, especially in data - scarce areas like the Arctic, generate realistic ship trajectory data to support the development of machine - learning (ML) models. Specifically, the authors use conditional generative adversarial networks (CGAN) to generate ship displacement trajectories from Automatic Identification System (AIS) data. In this way, the research aims to: 1. **Solve the data acquisition problem**: Due to privacy regulations and availability issues, it is often very difficult to obtain a large amount of high - quality real - world data. By generating synthetic data, this problem can be alleviated. 2. **Generate specific types of ship trajectories**: Generate trajectories of different types of ships (such as cargo ships and pleasure boats), especially in data - scarce situations (such as the Arctic region), or generate abnormal events for training detection models. 3. **Evaluate the quality of the generated data**: Verify whether the generated data can be effectively used for the training and testing of machine - learning models, ensuring its authenticity and practicality. ### Research Motivation and Contribution The main motivation of this research is to generate realistic ship trajectories for the complex coastline and ice conditions in the Canadian Arctic, and these trajectories will be used for testing defense and security monitoring algorithms. Specific contributions include: - **Apply GAN to geospatial ship motion data**: Generate ship trajectories in complex coastline areas. - **Use sparse input data to generate realistic ship positions**: Simulate situations similar to the Arctic region. - **Generate representative characteristics of two different types of ships**: Cargo ships and pleasure boats. - **Comprehensively evaluate model performance**: Evaluate through loss curves, column shapes, correlation measures, boundary compliance, and new row synthesis. - **Use machine - learning effectiveness tests**: Demonstrate the effectiveness of the generated data in actual deployment. - **Show the robustness of the model**: Generate data in different seasons to verify the stability of the model. ### Method Overview The research adopts conditional generative adversarial networks (CGAN), which contains two main components: - **Generator (G)**: Create synthetic data from random noise, imitating the distribution of real data. - **Discriminator (D)**: Receive real and synthetic data and estimate the probability that a data instance comes from the real - data category. CGAN can make the generator generate data that better meets the requirements according to specific conditions by introducing conditional information (such as ship type). The data set used in the experiment is from MarineCadastre.gov, covering AIS data in May, July, and September 2022. After data cleaning and pre - processing, the CGAN model is trained using a fully - connected feed - forward neural network architecture. ### Results and Discussion Through the evaluation of the generated data, the research finds that: - **The loss curves are stable**: The loss curves of the generator and the discriminator remain stable during the training process, indicating that the model converges well. - **The generated data is similar to the real data**: Evaluated by the Kolmogorov - Smirnov statistic, various features of the generated data are highly similar to the real data. - **Machine - learning effectiveness test**: The classification model trained with the generated data performs well on the real data, proving the effectiveness of the generated data. In conclusion, this research shows the potential of CGAN in generating realistic ship trajectories, especially in data - scarce situations, providing a powerful tool for marine environmental monitoring.

Generation of Vessel Track Characteristics Using a Conditional Generative Adversarial Network (CGAN)

Data-augmented Landslide Displacement Prediction Using Generative Adversarial Network

Conditional temporal GAN for intent-aware vessel trajectory prediction in the precautionary area

Generative Adversarial Network Models for Augmenting Digit and Character Datasets Embedded in Standard Markings on Ship Bodies

Unsupervised maritime anomaly detection for intelligent situational awareness using AIS data

A Novel Model for Ship Trajectory Anomaly Detection Based on Gaussian Mixture Variational Autoencoder

Data generation for connected and automated vehicle tests using deep learning models

Directional Wave Spectrum Estimation Using Ship Motion Data by Improved CGAN with Physics Guided

Conditional-TimeGAN for Realistic and High-Quality Appliance Trajectories Generation and Data Augmentation in Nonintrusive Load Monitoring

ShipHullGAN: A generic parametric modeller for ship hull design using deep convolutional generative model

Improvement of Machine Learning-Based Modelling of Container Ship's Main Particulars with Synthetic Data

Conditional generative adversarial networks for groundwater contamination characterization and source identification

Velocity Estimation of Ocean Surface Currents in along-Track InSAR System Based on Conditional Generative Adversarial Networks

Conditional GANs for Sonar Image Filtering with Applications to Underwater Occupancy Mapping

Regional ship behavior and trajectory prediction for maritime traffic management: A social generative adversarial network approach

Online Data-Driven Integrated Prediction Model for Ship Motion Based on Data Augmentation and Filtering Decomposition and Time-Varying Neural Network

A Prescriptive Model for Failure Analysis in Ship Machinery Monitoring Using Generative Adversarial Networks

A dual linear autoencoder approach for vessel trajectory prediction using historical AIS data

MTS-DVGAN: Anomaly Detection in Cyber-Physical Systems using a Dual Variational Generative Adversarial Network

Designing ship hull forms using generative adversarial networks

Recognition and Synthesis of Object Transport Motion