PaperlayerView exact revision on arXiv

Source-linked AI summary

A Large-Scale AIS Dataset from Finnish Water

Debayan Bhattacharya, Ikram Ul Haq, Carlos Pichardo Vicencio, Sebastien Lafond

arXiv:2609.12938v1cs.LG

TL;DR

The paper frames a need for high-quality, region-specific maritime datasets as maritime activities evolve. It surveys AIS datasets and presents a large dataset from Finnish waters, covering the Baltic Sea and Finnish lakes. The dataset contains 229 million data points, while the paper also visualizes traffic patterns and describes seasonal differences in activity.

  • Problem

    The paper frames a need for high-quality, region-specific datasets as maritime activities evolve and diversify.

    Source: §5 · p. 13

  • Method

    The paper surveys AIS datasets and presents a Finnish-waters AIS dataset spanning 20 months, from April 2021 to December 2022.

    Source: §1 · pp. 1–2, §2.1 · p. 2

  • Results

    The dataset contains 229 million data points, and its visualizations show seasonal differences in activity, including rarely observed lake activity during winter.

    Source: §1 · p. 1, §4.2 · p. 10

  • Takeaways & Limitations

    The dataset is presented as a resource for studying vessel behaviour, traffic patterns, and accident risk assessment in Finnish waters.

    Source: §5 · p. 13

    Some mapped AIS positions appear outside bodies of water, likely due to human errors in transmitter setup or GNSS receiver configuration.

    Source: §4.2 · p. 13

Abstract

from arXiv · show

This research paper contributes to the maritime research community by introducing a comprehensive AIS dataset from Finnish waters, specifically the Baltic Sea region. AIS data, initially designed for collision prevention, have evolved into a versatile tool with applications across diverse maritime domains. Our paper not only curates and categorises existing AIS datasets but also introduces a collected AIS dataset from the Baltic Sea area, renowned for its intercontinental cargo routes, military activities, and frozen water expanses. This dataset includes 229 millions data points and provides researchers with a resource for studying maritime activities and vessel behaviour in this dynamic region, notably distinguished by the inclusion of data from Finnish lakes. Our analysis includes a detailed listing of ship types and relevant features, empowering researchers to explore various maritime domains. To enhance comprehension and analysis, we provide visualisations of maritime traffic patterns. By publishing this AIS dataset, we aim to catalyse innovation and collaboration in maritime research, offering a gateway to deeper insights into maritime activities in the Baltic Sea and Finnish lakes.

1 Introduction

The paper presents a large AIS dataset focused on Finnish waters, covering Baltic Sea maritime activity and Finnish lakes. It also introduces AIS applications and visualizes traffic patterns.

1 Introduction

  • The dataset spans 20 months, from April 2021 to December 2022, and contains 229 million data points.It is presented as a resource for maritime research.Source: §1 · p. 1
  • AIS data are used across domains including navigation safety, maritime traffic monitoring, and collision avoidance.The paper offers visualizations that provide a glimpse of traffic patterns in its dataset.Source: §1 · pp. 1–2
  • The dataset examines vessel movements, trajectories, and contextual information in the Baltic Sea and Finnish lakes.The Baltic Sea is described as a region of intercontinental cargo exchange, military presence, and frozen-water challenges.Source: §1 · pp. 1–2

2 Use of AIS data in the literature

This section surveys machine-learning uses of AIS data and describes the kinds of maritime information AIS can provide. It also notes potential uses in historical analysis, prediction, and sensor-data fusion.

2 Use of AIS data in the literature

  • The section reviews machine-learning approaches using AIS data to summarize achievements and advances documented in the literature.The stated aim is to give readers an overview of these applications.Source: §2 · p. 2
  • AIS data can provide high- and low-level views of vessel particulars in a marine environment.The passage describes these data as supporting research across different domains.Source: §2 · p. 2
  • AIS data can be used to investigate historical events and train predictive models.The passage identifies both as potential research uses.Source: §2 · p. 2
  • Combining AIS with RADAR, LiDAR, or camera data is described as enabling a more comprehensive assessment of ship situational awareness.The combined data can be processed with AI and machine-learning algorithms.Source: §2 · p. 2

2.1 Methodology

The review used a Web of Science search centered on AIS and selected 35 relevant papers published before 2023.

2.1 Methodology

  • The review searched Web of Science for articles published before 2023, using “AIS” with “navigation,” “maritime,” and “machine learning” across all fields.The search identified 35 papers for analysis.Source: §2.1 · p. 2
  • The search centered on “AIS” and combined it with the keywords “navigation,” “maritime,” and “machine learning.”The keywords were applied across all fields in Web of Science.Source: §2.1 · p. 2
  • The review included articles published prior to 2023 and selected 35 papers for analysis.Table 1 presents the distribution of publication types.Source: §2.1 · p. 2

2.2 Results

The reviewed publication set spans several engineering and ocean disciplines, with China accounting for the largest reported share of publication origins. Publication volume rose to 11 papers in 2022.

2.2 Results

  • China accounted for 54 per cent of publication origins, while Norway and Singapore each accounted for 15 per cent.The most represented disciplines included Marine Engineering, Oceanography, Ocean Engineering, Electrical and Electronic Engineering, and Telecommunications Engineering.Source: §2.2 · p. 2
  • 11 publications appeared in 2022, compared with one each in 2016 and 2017.The paper describes this as an upward trend in publication volume.Source: §2.2 · p. 2

2.3 Application domains

The 33 retrieved publications span maritime application domains, with vessel trajectory determination most common and collision avoidance next. The section reviews work on collision avoidance, trajectory prediction, and ship-handling behavior patterns.

2.3 Application domains

  • Among 33 retrieved publications, vessel trajectory determination is the most common focus, followed by collision avoidance.Table 2 outlines the application domains and corresponding papers.Source: §2.3 · p. 2, Table 2 · p. 3

2.3.1 Collision Avoidance

  • Collision-avoidance research uses AIS to predict trajectories, identify potential crossing conflicts, and inform avoidance decisions.Approaches include encounter-decision learning, course prediction, and trajectory clustering for collision-scenario detection.Source: §2.3.1 · p. 4

2.3.2 Vessel Trajectory Prediction

  • Trajectory research uses AIS to predict and reconstruct vessel paths, cluster similar voyages, and support collision-risk assessment.Methods described include semi-supervised classification, clustering with DBSCAN and Gaussian mixture models, and recurrent models for reconstruction.Source: §2.3.2 · pp. 4–5

2.3.3 Ship handling behavior patterns

  • Ship-handling behavior research segments and clusters AIS trajectories to recognize movement patterns and identify anomalous behavior.One approach defines ship-handling behavior basic (SHBB) from clustered sub-trajectories; another uses DBSCAN to model behaviors from vessel profiles.Source: §2.3.3 · p. 5

2.3.4 Ship grounding risk assessment

This section surveys maritime applications including grounding-risk assessment, ship sensors, port-type prediction, risk analysis, anomaly detection, and accident analysis. The grounding-risk studies described include trajectory reconstruction and a spatial model for forecasting groundings.

2.3.4 Ship grounding risk assessment

  • The grounding model in achieved an R-squared value of 0.55 and a mean squared error of 0.002.The study forecasts both the occurrence and spatial pattern of groundings; another study reconstructed trajectories with Douglas-Peucker’s algorithm and clustered them by behavior.Source: §2.3.4 · p. 5

2.3.5 Ship Sensors

  • Autonomous-vessel research uses location, inertial, visual, audio, RADAR, and LiDAR sensors to collect data for vessel management.The passage notes that collected sensor data may also be displayed using augmented reality.Source: §2.3.5 · p. 5

2.3.6 Port Type Prediction

  • Port-type prediction has used supervised KNN and Random Forest classifiers.Source: §2.3.6 · p. 5

2.3.7 Risk Analysis

  • Maritime-risk analysis has used logistic regression followed by an artificial neural network, alongside datasets for different risk scenarios.Source: §2.3.7 · p. 6

2.3.8 Anomaly Detection

  • Anomaly-detection approaches include an ANN framework [33] and DBSCAN clustering of AIS point data.The section describes detecting deviations in vessel movements and traffic patterns using AIS, radar, and satellite imagery.Source: §2.3.8 · p. 6

2.3.9 Accident Analysis

  • AIS-based accident analysis examines vessel movements and contributing factors; one study built a data-informed Bayesian Network for significant incidents in China’s coastal waters.Source: §2.3.9 · p. 6

2.4 Other available AIS datasets

The section summarizes previously published AIS datasets and provides information about prior work. The datasets listed span free and paid options and include vessel positions, identifying details, and related features.

2.4 Other available AIS datasets

  • The section summarizes previously published AIS datasets, covering both free and paid options.The authors present the overview as information for researchers about prior work.Source: §2.4 · p. 6
  • The listed datasets provide ship positions, identifying details, and related features for research domains including transportation optimization and marine domain awareness.Source: §2.4 · p. 6

3 Proposed AIS dataset

The proposed AIS dataset spans 20 months across Finnish waters and contains more than 2.29 billion data points. It covers seas, rivers, and lakes and includes diverse ship types.

3 Proposed AIS dataset

  • The dataset covers April 2021 to December 2022 across seas, rivers, and lakes, with cargo ships, tankers, fishing vessels, passenger ships, and other categories.Its 20-month span supports analysis of long-term trends and potential influences such as weather, seasonal variations, and economic conditions.Source: §3 · p. 7
  • 2 293 129 345 data points provide granular records for examining vessel dynamics and operations in Finnish waters.The dataset also supports identifying patterns and anomalies in vessel behaviour and assessing potential environmental implications.Source: §3 · p. 7
  • The data were obtained through Fintraffic traffic services’ public MQTT WebSocket APIs and are organized as daily location and metadata CSV files.The dataset is freely available from.Source: §3 · p. 7

3.1 Details about location file

  • Location files contain received AIS position reports with timestamps, vessel identifiers, coordinates, motion, navigational status, and position-quality fields.Listed fields include MMSI, longitude, latitude, speed and course over ground, rate of turn, position accuracy, RAIM flag, and heading.Source: §3.1 · p. 7

3.2 Details about metadata file

  • Metadata files pair vessel identifiers with destination, call sign, IMO number, ship type, draught, ETA, positioning type, dimensions, and vessel name.The ETA provides information for planning and scheduling vessel movements.Source: §3.2 · pp. 8–9

4 Overview of the AIS dataset content

The dataset catalogs ship types and vessel identifiers, with distinctive coverage of Finnish lakes as well as Baltic waters. Heatmaps show seasonal shifts in activity, while also exposing positional inaccuracies in AIS data.

4 Overview of the AIS dataset content

  • Ship types use numeric codes from 1 to 99, while Table 4 and Table 5 report ship-type distributions by unique MMSI and IMO numbers, respectively.MMSI identifiers can change with ownership or flag-state registration; IMO numbers remain permanent throughout a ship’s lifetime.Source: §4 · p. 9, Table 4 · p. 10, Table 5 · p. 11

4.1 Ship features inside Finnish Lakes

  • Finnish-lake activity includes passenger ferries, cargo ships, research vessels, and rescue boats, with regular inland activity near Tampere, Jyväskylä, and Kuopio.The dataset’s lake coverage offers a perspective on vessel movements and navigational patterns distinct to Finnish lakes.Source: §4.1 · pp. 9–10
  • The illustrated Finnish territory covers selected lakes and rivers within the stated coordinate bounds, and Table 6 describes ship types across Finnish lakes by unique MMSI.The illustration uses data from June–August 2021 and 2022.Source: §4.1 · p. 10, Figure 3 · p. 11, Table 6 · p. 12

4.2 Visualisation of the AIS dataset content

  • Heatmaps aggregate ship activity by time period using Folium and a grid-search algorithm to reduce position points for efficient aggregation.The analysis examines monthly and seasonal periods, identifying concentrations of movement and prominent routes.Source: §4.2 · p. 10
  • Winter activity is concentrated around the Baltic Sea, while lake activity is rarely observed when frozen conditions prevent small-vessel use.The winter heatmap covers December 2021 to February 2022.Source: §4.2 · p. 10, Figure 4 · p. 12
  • Spring activity increases around ports including Helsinki and Tallinn, then spreads from the Baltic Sea toward nearby regions and shorelines.The spring heatmap covers March to May 2022.Source: §4.2 · p. 10, Figure 5 · p. 13
  • Summer brings increased activity across lakes and other water bodies, while autumn activity decreases and becomes more Baltic-centered, resembling spring.The summer heatmap covers June to August 2022; the autumn comparison is with summer and spring heatmaps.Source: §4.2 · p. 13
  • Some mapped AIS positions fall outside bodies of water, likely reflecting transmitter setup or GNSS receiver configuration errors.The authors describe these as real-world AIS positional inaccuracies.Source: §4.2 · p. 13

5 Conclusion

The paper introduces an extensive AIS dataset for Finnish waters, covering the Baltic Sea and surrounding lakes. The authors aim for it to support research on vessel behaviour, traffic patterns, and accident risk assessment.

5 Conclusion

  • The dataset focuses on Finnish waters in the Baltic Sea and surrounding lakes, with the stated aim of supporting research on vessel behaviour, traffic patterns, and accident risk assessment.The authors hope the resource will foster research collaboration and insights supporting safer, more efficient maritime operations.Source: §5 · p. 13

Funding

The work received partial support from the EDISS master’s programme through a European Union agency grant.

Funding

  • The work was partially supported by the EDISS master’s programme through European Union grant number 619819.Source: Funding · p. 14
Loading 2609.12938v1…