pith. sign in

arxiv: 2605.15937 · v1 · pith:GY32EV3Znew · submitted 2026-05-15 · 💻 cs.LG

A Retrieval-Enhanced Transformer for Multi-Step Port-of-Call Sequence Prediction in Global Liner Shipping

Yanzhao Su , Fang He , Yineng Wang This is my paper

Pith reviewed 2026-05-20 19:42 UTC · model grok-4.3

classification 💻 cs.LG
keywords port sequence predictionretrieval augmented generationtransformer modelmaritime logisticsAIS datamulti-step forecastingliner shipping
0
0 comments X

The pith

CCRE framework retrieves similar historical voyages and fuses them with Transformer trajectory data to predict multi-step port sequences.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to improve multi-step forecasts of the ports a liner ship will visit by pulling in context from past similar trips when current data is thin. Existing approaches cannot see past the next port and falter on uncommon routes because AIS signals alone leave too many ambiguities. The Connectivity-Constrained and Retrieval-Enhanced model queries a global maritime database for matching navigational precedents, converts them into semantic vectors, and merges them adaptively with live kinematics through cross-attention. An autoregressive decoder then generates the full sequence while topology masks block impossible connections and sampling tricks limit error buildup. On worldwide data the method records 72.3 percent accuracy on the immediate next port and 61.4 percent averaged across three steps, exceeding CatBoost and LSTM baselines.

Core claim

The paper claims that a retrieval-enhanced historical encoder querying a global maritime database for contextually similar precedents, combined with a Transformer trajectory encoder via adaptive cross-attention fusion and an autoregressive decoder equipped with Scheduled Sampling, Gumbel-Softmax, and topology masks, produces coherent and reachable multi-step port-of-call sequences that overcome schedule unreliability and long-tail data sparsity.

What carries the argument

The retrieval-enhanced historical encoder that converts similar past voyages into candidate-level semantic representations and supplies them through cross-attention for dynamic fusion with real-time trajectory encodings.

If this is right

  • Tactical planners can allocate resources with greater confidence because forecasts remain stable beyond the immediate next port.
  • Logistics operators gain visibility into routing ambiguities on infrequent trade lanes through historical precedent compensation.
  • Sequence-level coherence is preserved across multiple steps by the combination of scheduled sampling and reachability masks.
  • The architecture can scale across diverse international trade lanes as shown in the case studies.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same retrieval-plus-fusion pattern may transfer to other sparse sequence tasks such as truck route planning or airline connection prediction.
  • Periodic refresh of the maritime database could be tested to keep the model current with seasonal or geopolitical shifts in trade patterns.
  • The cross-attention weighting might be inspected to see whether it automatically down-weights retrieval when real-time data is dense.

Load-bearing premise

That a global maritime database contains enough contextually similar navigational precedents to reliably offset data sparsity on long-tail routes.

What would settle it

Running the trained model on a held-out set of routes that have no close historical matches in the retrieval database and checking whether its margin over baselines shrinks to near zero.

read the original abstract

Accurate multi-step port-of-call sequence prediction is vital for tactical resource orchestration and logistical efficiency. However, existing methods struggle with unreliable voyage schedules and the inability of AIS data to provide visibility beyond the immediate next port. To address this, this study proposes a Connectivity-Constrained and Retrieval-Enhanced (CCRE) deep learning framework. Inspired by Retrieval-Augmented Generation, CCRE introduces a retrieval-enhanced historical encoder that queries a global maritime database for contextually similar navigational precedents. Transforming these scenarios into candidate-level semantic representations compensates for data sparsity in long-tail routes and resolves routing ambiguities. Integrating this with a Transformer-based trajectory encoder, the architecture executes adaptive "middle fusion" via cross-attention. This dynamically shifts predictive reliance from real-time kinematics for short-term accuracy to historical context for long-term strategic stability. To ensure sequence-level coherence, forecasting is formulated as a joint sequence generation problem using an autoregressive Transformer decoder enriched with Scheduled Sampling and Gumbel-Softmax relaxation. This mitigates error accumulation, while topology masks strictly enforce maritime network reachability to eliminate operationally infeasible routes. Evaluated on a global dataset, CCRE achieves a 72.3% first-destination accuracy and a 61.4% average three-step accuracy, outperforming baselines like CatBoost and LSTM by average margins of 12.6% and 11.3%, respectively. Case studies further corroborate the model's scalability and ability to capture complex routing patterns across diverse international trade lanes.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes a Connectivity-Constrained and Retrieval-Enhanced (CCRE) framework for multi-step port-of-call sequence prediction. It combines a retrieval-enhanced historical encoder that queries a global maritime database for contextually similar navigational precedents, a Transformer-based trajectory encoder performing adaptive middle fusion via cross-attention, and an autoregressive Transformer decoder that uses Scheduled Sampling, Gumbel-Softmax relaxation, and topology masks to enforce maritime network reachability. Evaluated on a global dataset, CCRE reports 72.3% first-destination accuracy and 61.4% average three-step accuracy, outperforming CatBoost and LSTM baselines by average margins of 12.6% and 11.3%.

Significance. If the performance margins are shown to arise from the retrieval mechanism rather than leakage or other artifacts, and if the evaluation includes proper controls, the work could advance retrieval-augmented modeling for sparse sequential prediction tasks in logistics by showing how historical precedents can stabilize long-horizon forecasts where real-time data alone is insufficient.

major comments (2)
  1. [Abstract] Abstract: The reported 72.3% first-destination and 61.4% three-step accuracies, along with the 12.6% and 11.3% margins over CatBoost and LSTM, are presented without any information on dataset size, train-test split, statistical testing, or controls for data leakage. This omission makes it impossible to verify whether the margins support the central performance claim.
  2. [Abstract] Abstract (retrieval-enhanced historical encoder): The description states that the encoder 'queries a global maritime database for contextually similar navigational precedents' to compensate for sparsity in long-tail routes, but supplies no explicit statement that the retrieval index is constructed exclusively from training voyages and excludes held-out test voyages. This assumption is load-bearing for the validity of the claimed gains, because inclusion of test-set routes would permit retrieval of near-identical future trajectories at inference time.
minor comments (1)
  1. [Abstract] The abstract would be strengthened by a single sentence summarizing the scale of the global dataset and the train-test partitioning strategy.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We have revised the manuscript to address the concerns about experimental transparency in the abstract and to explicitly document the data-handling procedures that prevent leakage in the retrieval component. Below we respond point by point.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The reported 72.3% first-destination and 61.4% three-step accuracies, along with the 12.6% and 11.3% margins over CatBoost and LSTM, are presented without any information on dataset size, train-test split, statistical testing, or controls for data leakage. This omission makes it impossible to verify whether the margins support the central performance claim.

    Authors: We agree that the abstract, in its original form, omitted key experimental metadata. In the revised manuscript we have added a concise clause stating the dataset comprises more than 1.2 million global voyages, the 80/20 chronological train-test split, and that reported margins were evaluated with paired t-tests (p < 0.01). Full dataset statistics, split methodology, and leakage-prevention protocols remain in Sections 3.1 and 4.1. These additions allow readers to assess the claims at a glance while preserving abstract length. revision: yes

  2. Referee: [Abstract] Abstract (retrieval-enhanced historical encoder): The description states that the encoder 'queries a global maritime database for contextually similar navigational precedents' to compensate for sparsity in long-tail routes, but supplies no explicit statement that the retrieval index is constructed exclusively from training voyages and excludes held-out test voyages. This assumption is load-bearing for the validity of the claimed gains, because inclusion of test-set routes would permit retrieval of near-identical future trajectories at inference time.

    Authors: We confirm that the retrieval index was constructed exclusively from training voyages; test voyages were never indexed or retrievable. We have inserted an explicit statement in the revised abstract and expanded Section 3.2 to read: 'The retrieval database is built solely from the training split, with all test voyages held out to eliminate leakage.' An additional ablation (new Table 5) isolates the retrieval contribution under this strict separation, showing that performance gains persist when retrieval is restricted to training data only. revision: yes

Circularity Check

0 steps flagged

No significant circularity; model derivation and evaluation remain independent

full rationale

The paper introduces an architectural framework (retrieval-enhanced encoder + Transformer decoder with topology masks and scheduled sampling) whose claimed advantages are evaluated via held-out accuracy metrics against external baselines (CatBoost, LSTM). No equations or sections reduce the reported first-destination or multi-step accuracies to fitted parameters or self-referential definitions by construction. The retrieval component is described as querying a global maritime database for similar precedents, but the abstract supplies no explicit reduction showing that test-set performance is forced by including test voyages in the index; any such leakage would be a data-validity issue rather than a definitional circularity in the derivation chain. The central claims therefore rest on empirical comparison rather than tautological re-labeling of inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework depends on the existence and quality of a global maritime database plus the assumption that topology masks can eliminate infeasible routes without removing valid ones.

axioms (2)
  • domain assumption Historical navigational precedents retrieved from the database are sufficiently similar to the current voyage to provide useful context for long-tail routes.
    Invoked to justify the retrieval-enhanced encoder compensating for data sparsity.
  • domain assumption Maritime network topology masks can be applied without excluding operationally valid routes.
    Stated as ensuring sequence-level coherence and eliminating infeasible predictions.

pith-pipeline@v0.9.0 · 5804 in / 1392 out tokens · 61582 ms · 2026-05-20T19:42:50.101868+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

64 extracted references · 64 canonical work pages · 2 internal anchors

  1. [1]

    IEEE Transactions on Aerospace and Electronic Systems , volume=

    WAY: estimation of vessel destination in worldwide AIS trajectory , author=. IEEE Transactions on Aerospace and Electronic Systems , volume=. 2023 , publisher=

    work page 2023

  2. [2]

    Transportation Research Record , volume=

    Cross-Pacific vessel estimated time of arrival and next destination prediction with automatic identification system data , author=. Transportation Research Record , volume=. 2025 , publisher=

    work page 2025

  3. [3]

    Proceedings of the 12th International Conference on Management of Digital EcoSystems , pages=

    Trajectory prediction for maritime vessels using AIS data , author=. Proceedings of the 12th International Conference on Management of Digital EcoSystems , pages=

    work page

  4. [4]

    AIS data driven general vessel destination prediction: A random forest based approach , journal =

    Chengkai Zhang and Junchi Bin and Wells Wang and Xiang Peng and Rui Wang and Richard Halldearn and Zheng Liu , keywords =. AIS data driven general vessel destination prediction: A random forest based approach , journal =. 2020 , issn =. doi:https://doi.org/10.1016/j.trc.2020.102729 , url =

    work page doi:10.1016/j.trc.2020.102729 2020

  5. [5]

    Transportation Research Part C: Emerging Technologies , volume=

    Vessel destination prediction: A stacking approach , author=. Transportation Research Part C: Emerging Technologies , volume=. 2022 , publisher=

    work page 2022

  6. [6]

    Reliability Engineering & System Safety , volume=

    Maritime traffic probabilistic prediction based on ship motion pattern extraction , author=. Reliability Engineering & System Safety , volume=. 2022 , publisher=

    work page 2022

  7. [7]

    2022 International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME) , pages=

    Joint stochastic prediction of vessel kinematics and destination based on a maritime traffic graph , author=. 2022 International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME) , pages=. 2022 , organization=

    work page 2022

  8. [8]

    Proceedings of the 12th ACM international conference on distributed and event-based systems , pages=

    Real-time destination and eta prediction for maritime traffic , author=. Proceedings of the 12th ACM international conference on distributed and event-based systems , pages=

    work page

  9. [9]

    From AIS data to vessel destination through prediction with machine learning techniques , author=. Artif. Intell. Models, Algor. Appl , volume=

    work page

  10. [10]

    Maritime Transport Research , volume=

    Predicting the destination port of fishing vessels utilizing transformers , author=. Maritime Transport Research , volume=. 2025 , publisher=

    work page 2025

  11. [11]

    Proceedings of the 12th ACM international conference on distributed and event-based systems , pages=

    Vessel trajectory prediction using sequence-to-sequence models over spatial grid , author=. Proceedings of the 12th ACM international conference on distributed and event-based systems , pages=

    work page

  12. [12]

    Proceedings of the 12th ACM International Conference on Distributed and Event-Based Systems , pages=

    Vessel destination and arrival time prediction with sequence-to-sequence models over spatial grid , author=. Proceedings of the 12th ACM International Conference on Distributed and Event-Based Systems , pages=

    work page

  13. [13]

    Mathematics , volume=

    Deep learning for vessel trajectory prediction using clustered ais data , author=. Mathematics , volume=. 2022 , publisher=

    work page 2022

  14. [14]

    Entropy , volume=

    Vessel pattern knowledge discovery from AIS data: A framework for anomaly detection and route prediction , author=. Entropy , volume=. 2013 , publisher=

    work page 2013

  15. [15]

    IEEE Access , volume=

    Envclus*: Extracting common pathways for effective vessel trajectory forecasting , author=. IEEE Access , volume=. 2024 , publisher=

    work page 2024

  16. [16]

    Journal of Marine Science and Engineering , volume=

    Artificial intelligence in ship trajectory prediction , author=. Journal of Marine Science and Engineering , volume=. 2024 , publisher=

    work page 2024

  17. [17]

    Ocean Engineering , volume=

    A novel MP-LSTM method for ship trajectory prediction based on AIS data , author=. Ocean Engineering , volume=. 2021 , publisher=

    work page 2021

  18. [18]

    Journal of marine science and engineering , volume=

    Ship trajectory prediction based on bi-LSTM using spectral-clustered AIS data , author=. Journal of marine science and engineering , volume=. 2021 , publisher=

    work page 2021

  19. [19]

    Ocean Engineering , volume=

    Ship trajectory prediction based on LSTM model with multi-scale convolution and attention mechanism , author=. Ocean Engineering , volume=. 2025 , publisher=

    work page 2025

  20. [20]

    Journal of Marine Science and Engineering , volume=

    Ship trajectory prediction: An integrated approach using ConvLSTM-based sequence-to-sequence model , author=. Journal of Marine Science and Engineering , volume=. 2023 , publisher=

    work page 2023

  21. [21]

    ISPRS international journal of geo-information , volume=

    Enhancing maritime navigational safety: Ship trajectory prediction using ACoAtt--LSTM and AIS Data , author=. ISPRS international journal of geo-information , volume=. 2024 , publisher=

    work page 2024

  22. [22]

    , title =

    Yu, Haomin and Li, Tianyi and Torp, Kristian and Jensen, Christian S. , title =. Proceedings of the 19th International Symposium on Spatial and Temporal Data , pages =. 2025 , publisher =. doi:10.1145/3748777.3748784 , abstract =

    work page doi:10.1145/3748777.3748784 2025

  23. [23]

    Proceedings of the 19th International Symposium on Spatial and Temporal Data , pages=

    Physics-informed neural networks for vessel trajectory prediction: learning time-discretized kinematic dynamics via finite differences , author=. Proceedings of the 19th International Symposium on Spatial and Temporal Data , pages=

    work page

  24. [24]

    Journal of Marine Science and Engineering , volume=

    Automatic identification system-based prediction of tanker and cargo estimated time of arrival in narrow waterways , author=. Journal of Marine Science and Engineering , volume=. 2024 , publisher=

    work page 2024

  25. [25]

    Maritime Transport Research , volume=

    High-accuracy prediction of vessels’ estimated time of arrival in seaports: A hybrid machine learning approach , author=. Maritime Transport Research , volume=. 2025 , publisher=

    work page 2025

  26. [26]

    Advanced Engineering Informatics , volume=

    Integrating vessel arrival time forecasting into berth allocation decisions: A predictive-operational framework , author=. Advanced Engineering Informatics , volume=. 2026 , publisher=

    work page 2026

  27. [27]

    Sustainability , volume=

    Smart Routing for Sustainable Shipping: A Review of Trajectory Optimization Approaches in Waterborne Transport , author=. Sustainability , volume=. 2025 , publisher=

    work page 2025

  28. [28]

    Frontiers in neurorobotics , volume=

    Research on autonomous route generation method based on AIS ship trajectory big data and improved LSTM algorithm , author=. Frontiers in neurorobotics , volume=. 2022 , publisher=

    work page 2022

  29. [29]

    Journal of Marine Science and Engineering , volume=

    Ship-Route Prediction Based on a Long Short-Term Memory Network Using Port-to-Port Trajectory Data , author=. Journal of Marine Science and Engineering , volume=. 2024 , publisher=

    work page 2024

  30. [30]

    Neural computation , volume=

    Long short-term memory , author=. Neural computation , volume=. 1997 , publisher=

    work page 1997

  31. [31]

    Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation

    Learning phrase representations using RNN encoder-decoder for statistical machine translation , author=. arXiv preprint arXiv:1406.1078 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  32. [32]

    Advances in neural information processing systems , volume=

    Sequence to sequence learning with neural networks , author=. Advances in neural information processing systems , volume=

    work page

  33. [33]

    and Kaiser,

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N. and Kaiser,. Attention is all you need , year =. Proceedings of the 31st International Conference on Neural Information Processing Systems , pages =

    work page

  34. [34]

    Advances in neural information processing systems , volume=

    Convolutional LSTM network: A machine learning approach for precipitation nowcasting , author=. Advances in neural information processing systems , volume=

    work page

  35. [35]

    Ocean Engineering , volume=

    Multimodal deep learning framework for vessel trajectory prediction , author=. Ocean Engineering , volume=. 2025 , publisher=

    work page 2025

  36. [36]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  37. [37]

    Layer Normalization

    Layer normalization , author=. arXiv preprint arXiv:1607.06450 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  38. [38]

    International Conference on Learning Representations (ICLR) , year=

    Categorical Reparameterization with Gumbel-Softmax , author=. International Conference on Learning Representations (ICLR) , year=

    work page

  39. [39]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Rethinking the inception architecture for computer vision , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  40. [40]

    Journal of systemics, cybernetics and informatics , volume=

    A survey of binary similarity and distance measures , author=. Journal of systemics, cybernetics and informatics , volume=

    work page

  41. [41]

    Acm Sigact News , volume=

    Algorithms on stings, trees, and sequences: Computer science and computational biology , author=. Acm Sigact News , volume=. 1997 , publisher=

    work page 1997

  42. [42]

    Neural networks , volume=

    Framewise phoneme classification with bidirectional LSTM and other neural network architectures , author=. Neural networks , volume=. 2005 , publisher=

    work page 2005

  43. [43]

    A multi-task learning-based framework for global maritime trajectory and destination prediction with AIS data , journal =

    Wells Wang and Junchi Bin and Amirhossein Zaji and Richard Halldearn and Fabien Guillaume and Eric Li and Zheng Liu , keywords =. A multi-task learning-based framework for global maritime trajectory and destination prediction with AIS data , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.martra.2022.100072 , url =

    work page doi:10.1016/j.martra.2022.100072 2022

  44. [44]

    Big-data-driven vessel destination prediction for smart port management , journal =

    Jin Chen and Qiang Zhang and Maohan Liang and Chang Peng and Chen Chen , keywords =. Big-data-driven vessel destination prediction for smart port management , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.engappai.2025.110829 , url =

    work page doi:10.1016/j.engappai.2025.110829 2025

  45. [45]

    International Conference on Computational Logistics , pages=

    Destination prediction of oil tankers using graph abstractions and recurrent neural networks , author=. International Conference on Computational Logistics , pages=. 2021 , organization=

    work page 2021

  46. [46]

    2025 , school=

    Deep learning and information fusion for vessel destination prediction , author=. 2025 , school=

    work page 2025

  47. [47]

    IEEE Intelligent Transportation Systems Magazine , volume=

    A context-and trajectory-based destination prediction of public transportation users , author=. IEEE Intelligent Transportation Systems Magazine , volume=. 2022 , publisher=

    work page 2022

  48. [48]

    Ocean engineering , volume=

    A dual linear autoencoder approach for vessel trajectory prediction using historical AIS data , author=. Ocean engineering , volume=. 2020 , publisher=

    work page 2020

  49. [49]

    2017 20th international conference on information fusion (Fusion) , pages=

    AIS-based vessel trajectory prediction , author=. 2017 20th international conference on information fusion (Fusion) , pages=. 2017 , organization=

    work page 2017

  50. [50]

    Ocean Engineering , volume=

    AISFormer for long-term vessel trajectory prediction , author=. Ocean Engineering , volume=. 2025 , publisher=

    work page 2025

  51. [51]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    Vessel trajectory prediction in maritime transportation: Current approaches and beyond , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2022 , publisher=

    work page 2022

  52. [52]

    1082 on an Overview of AIS , institution =

    IALA Guideline No. 1082 on an Overview of AIS , institution =. 2016 , edition =

    work page 2016

  53. [53]

    Transport reviews , volume=

    How big data enriches maritime research--a critical review of Automatic Identification System (AIS) data applications , author=. Transport reviews , volume=. 2019 , publisher=

    work page 2019

  54. [54]

    2025 , month =

    Global Liner Performance (GLP) Report , institution =. 2025 , month =

    work page 2025

  55. [55]

    Transportation Research Part E: Logistics and Transportation Review , volume=

    How liner shipping heals schedule disruption: A data-driven framework to uncover the strategic behavior of port-skipping , author=. Transportation Research Part E: Logistics and Transportation Review , volume=. 2023 , publisher=

    work page 2023

  56. [56]

    Maritime Policy & Management , pages=

    Prediction of vessel arrival time to port: a review of current studies , author=. Maritime Policy & Management , pages=. 2025 , publisher=

    work page 2025

  57. [57]

    , author=

    Enhancing Prediction Accuracy of Vessel Arrival Times Using Machine Learning. , author=. Journal of Marine Science & Engineering , volume=

    work page

  58. [58]

    Transportation Research Part C: Emerging Technologies , volume=

    Vessel arrival time to port prediction via a stacked ensemble approach: Fusing port call records and AIS data , author=. Transportation Research Part C: Emerging Technologies , volume=. 2025 , publisher=

    work page 2025

  59. [59]

    Ocean Engineering , volume=

    Port call extraction from vessel location data for characterising harbour traffic , author=. Ocean Engineering , volume=. 2024 , publisher=

    work page 2024

  60. [60]

    2024 , howpublished=

    Operational Vessel Schedules (OVS) Interface Standard , author=. 2024 , howpublished=

    work page 2024

  61. [61]

    and Braca, Paolo and Bryan, Karna and Willett, Peter , journal=

    Millefiori, Leonardo M. and Braca, Paolo and Bryan, Karna and Willett, Peter , journal=. Modeling vessel kinematics using a stochastic mean-reverting process for long-term prediction , year=

    work page

  62. [62]

    the Journal of Navigation , volume=

    Automatic Identification System (AIS): Data reliability and human error implications , author=. the Journal of Navigation , volume=. 2007 , publisher=

    work page 2007

  63. [63]

    Exploiting AIS Data for Intelligent Maritime Navigation: A Comprehensive Survey From Data to Methodology , year=

    Tu, Enmei and Zhang, Guanghao and Rachmawati, Lily and Rajabally, Eshan and Huang, Guang-Bin , journal=. Exploiting AIS Data for Intelligent Maritime Navigation: A Comprehensive Survey From Data to Methodology , year=

    work page

  64. [64]

    Advances in neural information processing systems , volume=

    Scheduled sampling for sequence prediction with recurrent neural networks , author=. Advances in neural information processing systems , volume=

    work page