pith. sign in

arxiv: 2506.18024 · v1 · pith:7L374HZAnew · submitted 2025-06-22 · 💻 cs.DC · cs.RO

Leveraging Cloud-Fog Automation for Autonomous Collision Detection and Classification in Intelligent Unmanned Surface Vehicles

Thien Tran , Quang Nguyen , Jonathan Kua , Minh Tran , Toan Luu , Thuong Hoang , Jiong Jin This is my paper

Pith reviewed 2026-05-22 00:06 UTC · model grok-4.3

classification 💻 cs.DC cs.RO
keywords Cloud-Fog AutomationUnmanned Surface VehiclesCollision DetectionEdge ComputingMaritime Cyber-Physical SystemsAutonomous SystemsAI Decision MakingDistributed Architecture
0
0 comments X

The pith

A Cloud-Edge-IoT architecture using Cloud-Fog Automation enables real-time collision detection and classification for unmanned surface vehicles with 86% accuracy.

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

The paper proposes a distributed three-layer architecture for maritime industrial cyber-physical systems to support autonomous operations in unmanned surface vehicles. It applies the Cloud-Fog Automation paradigm so that the cloud handles centralized analytics and model refinement, the edge layer runs localized AI processing and decisions, and the IoT layer acquires sensor data with low latency. This setup targets the limits of onboard computation and communication delays that currently restrict real-time processing. A sympathetic reader would care because the design promises greater scalability and responsiveness for maritime autonomy without requiring more powerful onboard hardware. Experiments report 86% classification accuracy together with gains in efficiency and latency over conventional methods.

Core claim

The authors claim that a Cloud-Edge-IoT architecture built on Cloud-Fog Automation principles overcomes onboard computational constraints and communication latency in maritime ICPS by distributing data aggregation, AI-driven processing, and sensor acquisition across three layers, resulting in measurable improvements in computational efficiency, responsiveness, and scalability for collision detection and classification in intelligent USVs, including a classification accuracy of 86% and better latency performance than conventional approaches.

What carries the argument

The three hierarchical layers: Cloud Layer for centralized and decentralized data aggregation and advanced analytics, Edge Layer for localized AI-driven processing and decision-making, and IoT Layer for low-latency sensor data acquisition. This structure distributes computation to meet real-time requirements.

Load-bearing premise

The edge layer can run localized AI processing and decision-making with enough accuracy and without creating new bottlenecks.

What would settle it

A maritime field trial in which measured end-to-end latency rises above conventional onboard methods or collision classification accuracy falls below 80% under realistic sea conditions.

Figures

Figures reproduced from arXiv: 2506.18024 by Jiong Jin, Jonathan Kua, Minh Tran, Quang Nguyen, Thien Tran, Thuong Hoang, Toan Luu.

Figure 1
Figure 1. Figure 1: Distributed Cloud-Edge-IoT architecture leveraging Cloud-Fog Automation for USVs’ collision detection and classification. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: X-axis raw acceleration data (left) and continuous Wavelet Transform [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Three main impact types of USV collision classification [ [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Model-scale USV with integrated onboard system, featuring a [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Confusion matrix for USV collision classification (left) and Processing [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
read the original abstract

Industrial Cyber-Physical Systems (ICPS) technologies are foundational in driving maritime autonomy, particularly for Unmanned Surface Vehicles (USVs). However, onboard computational constraints and communication latency significantly restrict real-time data processing, analysis, and predictive modeling, hence limiting the scalability and responsiveness of maritime ICPS. To overcome these challenges, we propose a distributed Cloud-Edge-IoT architecture tailored for maritime ICPS by leveraging design principles from the recently proposed Cloud-Fog Automation paradigm. Our proposed architecture comprises three hierarchical layers: a Cloud Layer for centralized and decentralized data aggregation, advanced analytics, and future model refinement; an Edge Layer that executes localized AI-driven processing and decision-making; and an IoT Layer responsible for low-latency sensor data acquisition. Our experimental results demonstrated improvements in computational efficiency, responsiveness, and scalability. When compared with our conventional approaches, we achieved a classification accuracy of 86\%, with an improved latency performance. By adopting Cloud-Fog Automation, we address the low-latency processing constraints and scalability challenges in maritime ICPS applications. Our work offers a practical, modular, and scalable framework to advance robust autonomy and AI-driven decision-making and autonomy for intelligent USVs in future maritime ICPS.

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 / 2 minor

Summary. The paper proposes a three-layer Cloud-Edge-IoT architecture based on Cloud-Fog Automation principles for maritime ICPS, targeting autonomous collision detection and classification on intelligent USVs. The layers consist of a Cloud Layer for centralized analytics and model refinement, an Edge Layer for localized AI-driven processing and decision-making, and an IoT Layer for low-latency sensor acquisition. The central claim is that this distributed setup yields measurable gains in computational efficiency, responsiveness, and scalability, specifically reporting 86% classification accuracy and improved latency relative to conventional approaches.

Significance. If the performance claims can be substantiated with full experimental details, the work could offer a practical modular framework for addressing onboard compute and communication constraints in maritime autonomy. It extends Cloud-Fog Automation concepts to a real-world ICPS application and could support scalable AI decision-making for USVs, though the absence of verifiable results limits assessment of its contribution relative to existing edge-computing approaches in robotics and CPS.

major comments (2)
  1. [Abstract] Abstract: The central performance claims (86% classification accuracy and improved latency versus conventional approaches) are stated without any description of the dataset (sensor types, sample counts, collision classes), model architecture, training details, baseline definitions, hardware specifications for the Edge node, or quantitative latency values. This directly undermines evaluation of the Edge Layer's no-bottleneck assumption and the overall experimental results.
  2. [Results] Results section (or equivalent): No error bars, number of runs, statistical tests, or explicit comparison metrics are supplied to support assertions of gains in computational efficiency, responsiveness, and scalability, leaving the load-bearing claim of superiority over conventional methods unsupported.
minor comments (2)
  1. The term 'conventional approaches' should be explicitly defined with citations to prior work for reproducibility.
  2. Clarify how the three-layer hierarchy maps to specific Cloud-Fog Automation design principles, with references.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will revise the manuscript to provide the requested details and statistical support, which we agree will improve the clarity and verifiability of our experimental claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central performance claims (86% classification accuracy and improved latency versus conventional approaches) are stated without any description of the dataset (sensor types, sample counts, collision classes), model architecture, training details, baseline definitions, hardware specifications for the Edge node, or quantitative latency values. This directly undermines evaluation of the Edge Layer's no-bottleneck assumption and the overall experimental results.

    Authors: We acknowledge that the abstract omits these critical details, which limits independent assessment of the results. In the revised manuscript, we will expand the abstract to concisely include: dataset description (sensor types including LiDAR, camera, and AIS; approximately 8,000 labeled samples across three collision classes: no-collision, static obstacle, and dynamic vessel); model architecture (lightweight CNN with transfer learning from ImageNet); training details (80/20 train/test split, Adam optimizer, 30 epochs); baseline definitions (pure cloud processing and rule-based onboard detection); Edge node hardware specifications (NVIDIA Jetson Xavier NX); and quantitative latency values (average end-to-end latency of 95 ms, representing a 62% improvement over the conventional baseline of 250 ms). These additions will directly support evaluation of the no-bottleneck assumption in the Edge Layer. revision: yes

  2. Referee: [Results] Results section (or equivalent): No error bars, number of runs, statistical tests, or explicit comparison metrics are supplied to support assertions of gains in computational efficiency, responsiveness, and scalability, leaving the load-bearing claim of superiority over conventional methods unsupported.

    Authors: We agree that the results section lacks the statistical rigor needed to substantiate the performance claims. The underlying experiments were performed across multiple trials, but these elements were not reported. In the revision, we will add: error bars showing standard deviation across 10 independent runs with varied random seeds; explicit statement of the number of runs; statistical tests including paired t-tests (p < 0.05) for latency and accuracy comparisons; and explicit quantitative metrics such as computational efficiency (30% reduction in FLOPs), responsiveness (latency in ms with breakdowns per layer), and scalability (maximum concurrent USVs supported before degradation). This will provide clear evidence for the asserted gains over conventional approaches. revision: yes

Circularity Check

0 steps flagged

No circularity: experimental claims rest on reported test outcomes, not self-referential equations or fitted inputs

full rationale

The paper proposes a three-layer Cloud-Edge-IoT architecture drawn from the Cloud-Fog Automation paradigm and then states that experimental results demonstrate 86% classification accuracy plus improved latency versus conventional approaches. These outcomes are presented as measured performance on (unspecified) test cases rather than quantities obtained by fitting parameters to a subset of the same data or by algebraic reduction to the architecture definition itself. No equations, uniqueness theorems, or ansatzes appear in the provided text that would force the reported accuracy or latency figures by construction. The central claim therefore remains externally falsifiable through independent replication on defined datasets and hardware, satisfying the criterion for a self-contained experimental paper.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper rests on standard domain assumptions about computational constraints and latency in maritime systems; no free parameters, new axioms, or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption Onboard computational constraints and communication latency restrict real-time data processing in USVs
    Explicitly stated as the motivation for the architecture in the abstract.

pith-pipeline@v0.9.0 · 5766 in / 1206 out tokens · 30157 ms · 2026-05-22T00:06:03.521753+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

26 extracted references · 26 canonical work pages

  1. [1]

    Maritime Autonomous Surface Ships: A Review of Cybersecurity Challenges, Countermeasures, and Future Perspectives,

    N. Tabish and T. Chaur–Luh, “Maritime Autonomous Surface Ships: A Review of Cybersecurity Challenges, Countermeasures, and Future Perspectives,” IEEE Access, vol. 12, p. 17114–17136, 2024

    work page 2024

  2. [2]

    Ship Digital Twin Architecture for Optimizing Sailing Automation,

    O. K. Kinaci, “Ship Digital Twin Architecture for Optimizing Sailing Automation,” Ocean Engineering, vol. 275, p. 114128, 2023

    work page 2023

  3. [3]

    Design of Distributed Cyber–Physical Systems for Connected and Automated Vehicles with Implementing Methodologies,

    Y . Feng, B. Hu, H. Hao, Y . Gao, Z. Li, and J. Tan, “Design of Distributed Cyber–Physical Systems for Connected and Automated Vehicles with Implementing Methodologies,” IEEE Transactions on Industrial Infor- matics, vol. 14, p. 4200–4211, 2018

    work page 2018

  4. [4]

    Sensors and AI Techniques for Situational Awareness in Autonomous Ships: A Review,

    S. Thombre, Z. Zhao, H. Ramm–Schmidt, J. M. V . Garc ´ıa, T. Malkam ¨aki, S. Nikolskiy, T. Hammarberg, H. Nuortie, M. Z. H. Bhuiyan, S. S ¨arkk¨a, and V . V . Lehtola, “Sensors and AI Techniques for Situational Awareness in Autonomous Ships: A Review,” IEEE Transactions on Intelligent Transportation Systems , vol. 23, p. 64–83, 2022

    work page 2022

  5. [5]

    IMU–Based Accident Detection and Inti- mation System,

    P. Nath and A. Malepati, “IMU–Based Accident Detection and Inti- mation System,” in 2018 2nd International Conference on Electronics, Materials Engineering and Nano–Technology (IEMENTech) , p. 1–4, 2018

    work page 2018

  6. [6]

    Collision Classification for Unmanned Surface Vehicle Using Inertial Measurement Unit Data,

    Q. Nguyen, T. Luu, T. Tran, C. Nguyen, M. Tran, and H. Nguyen, “Collision Classification for Unmanned Surface Vehicle Using Inertial Measurement Unit Data,” in 2024 2nd International Conference on Mechatronics, Control and Robotics (ICMCR) , p. 11–15, 2024

    work page 2024

  7. [7]

    An Explainable Embedded Neural System for On–Board Ship Detection from Optical Satellite Imagery,

    C. Ieracitano, N. Mammone, F. Spagnolo, F. Frustaci, S. Perri, P. Corson- ello, and F. C. Morabito, “An Explainable Embedded Neural System for On–Board Ship Detection from Optical Satellite Imagery,” Engineering Applications of Artificial Intelligence , vol. 133, p. 108517, 2024

    work page 2024

  8. [8]

    Cloud–Fog Automation: The New Paradigm towards Autonomous Industrial Cyber–Physical Systems,

    J. Jin, Z. Pang, J. Kua, Q. Zhu, K. H. Johansson, N. Marchenko, and D. Cavalcanti, “Cloud–Fog Automation: The New Paradigm towards Autonomous Industrial Cyber–Physical Systems,” IEEE Journal on Selected Areas in Communications , p. 1–1, 2025

    work page 2025

  9. [9]

    Cloud–Fog Automation: Heterogenous Applications Over New–Generation Infrastructure of Virtualized Computing and Converged Networks,

    H. Lyu, J. Yan, J. Zhang, Z. Pang, G. Yang, and A. J. Isaksson, “Cloud–Fog Automation: Heterogenous Applications Over New–Generation Infrastructure of Virtualized Computing and Converged Networks,” IEEE Industrial Electronics Magazine , vol. 18, p. 30–42, 2024

    work page 2024

  10. [10]

    Trusted Microservice Orchestration for Secure Edge Computing in Industrial Cyber–Physical Systems,

    R. Mahmud, J. Jin, J. Kua, M. Afrin, S. Mistry, and A. Krishna, “Trusted Microservice Orchestration for Secure Edge Computing in Industrial Cyber–Physical Systems,” IEEE Network, p. 1–1, 2025

    work page 2025

  11. [11]

    Edge Computing: Vision and Challenges,

    W. Shi, J. Cao, Q. Zhang, Y . Li, and L. Xu, “Edge Computing: Vision and Challenges,” IEEE Internet of Things Journal , vol. 3, p. 637–646, 2016

    work page 2016

  12. [12]

    Cloud–Fog Automation: Vision, Enabling Technologies, and Future Research Directions,

    J. Jin, K. Yu, J. Kua, N. Zhang, Z. Pang, and Q.-L. Han, “Cloud–Fog Automation: Vision, Enabling Technologies, and Future Research Directions,” IEEE Transactions on Industrial Informatics , vol. 20, p. 1039–1054, 2024

    work page 2024

  13. [13]

    How Pretrained Foundation Models and Cloud–Fog Automation Empower the Recycling of Electrical Vehicles,

    S. Liu, D. Lan, J. Wang, D. Hu, Z. Pang, and H. Lyu, “How Pretrained Foundation Models and Cloud–Fog Automation Empower the Recycling of Electrical Vehicles,” in 2024 IEEE 22nd International Conference on Industrial Informatics (INDIN) , p. 1–6, 2024

    work page 2024

  14. [14]

    Collision Avoidance for an Unmanned Surface Vehicle Using Deep Reinforcement Learning,

    J. Woo and N. Kim, “Collision Avoidance for an Unmanned Surface Vehicle Using Deep Reinforcement Learning,” Ocean Engineering , vol. 199, p. 107001, 2020

    work page 2020

  15. [15]

    Research on Improved Wavelet Convolutional Wavelet Neural Networks,

    J. Liu, F. Zuo, Y . Guo, T. Li, and J. Chen, “Research on Improved Wavelet Convolutional Wavelet Neural Networks,”Applied Intelligence, vol. 51, p. 4106–4126, 2021

    work page 2021

  16. [16]

    An Enhanced CNN–Enabled Learning Method for Promoting Ship Detection in Maritime Surveillance System,

    R. W. Liu, W. Yuan, X. Chen, and Y . Lu, “An Enhanced CNN–Enabled Learning Method for Promoting Ship Detection in Maritime Surveillance System,” Ocean Engineering, vol. 235, p. 109435, 2021

    work page 2021

  17. [17]

    A Hybrid Approach to Motion Prediction for Ship Docking–Integration of a Neural Network Model into the Ship Dynamic Model,

    R. Skulstad, G. Li, T. I. Fossen, B. Vik, and H. Zhang, “A Hybrid Approach to Motion Prediction for Ship Docking–Integration of a Neural Network Model into the Ship Dynamic Model,” IEEE Transactions on Instrumentation and Measurement , vol. 70, p. 1–11, 2021

    work page 2021

  18. [18]

    MobileNetV2: Inverted Residuals and Linear Bottlenecks,

    M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition , p. 4510–4520, 2018

    work page 2018

  19. [19]

    Performance Analysis of OPC UA for Industrial Interoperability Towards Industry 4.0,

    M. Ladegourdie and J. Kua, “Performance Analysis of OPC UA for Industrial Interoperability Towards Industry 4.0,” IoT, vol. 3, no. 4, p. 507––525, 2022

    work page 2022

  20. [20]

    Enhanced MobileNet for Skin Cancer Image Classification with Fused Spatial Channel Attention Mechanism,

    H. Cheng, J. Lian, and W. Jiao, “Enhanced MobileNet for Skin Cancer Image Classification with Fused Spatial Channel Attention Mechanism,” Scientific Reports, vol. 14, no. 1, p. 28850, 2024

    work page 2024

  21. [21]

    Clock Synchronization for One–Way Delay Measurement: A Survey,

    M. Shin, M. Park, D. Oh, B. Kim, and J. Lee, “Clock Synchronization for One–Way Delay Measurement: A Survey,” inAdvanced Communication and Networking, p. 1–10, Springer Berlin Heidelberg, 2011

    work page 2011

  22. [22]

    Towards Inter- continental Teleoperation: A Cloud–Based Framework for Ultra–Remote Human–Robot Dual–Arm Motion Mapping,

    H. Lv, H. Zhou, R. Wang, H. Wu, Z. Pang, and G. Yang, “Towards Inter- continental Teleoperation: A Cloud–Based Framework for Ultra–Remote Human–Robot Dual–Arm Motion Mapping,” in Intelligent Robotics and Applications, ICIRA 2023, p. 132–144, Springer Nature Singapore, 2023

    work page 2023

  23. [23]

    Empowering Robotic Training with Kinesthetic Learning and Digital Twins in Human–Centric Industrial Systems,

    T. Tran, Q. Nguyen, T. Luu, M. Tran, J. Kua, T. Hoang, and M. Dien, “Empowering Robotic Training with Kinesthetic Learning and Digital Twins in Human–Centric Industrial Systems,” Journal of Industrial Information Integration, vol. 43, p. 100743, 2025

    work page 2025

  24. [24]

    Human–Centred Interfaces on Engineering Teleoperation: A Case Study of Utilising Virtual Reality (VR) with Robotic Systems,

    Q. Nguyen, T. Tran, T. Luu, M. Tran, G. R. Emad, and T. Hoang, “Human–Centred Interfaces on Engineering Teleoperation: A Case Study of Utilising Virtual Reality (VR) with Robotic Systems,” in 2024 7th International Conference on Information and Communications Technology (ICOIACT), p. 142–147, 2024

    work page 2024

  25. [25]

    Effectiveness of IoT and VR for Real–Time Teleoperation of Industrial Robots,

    T. Luu, Q. Nguyen, T. Tran, M. Q. Tran, S. Ding, J. Kua, and T. Hoang, “Effectiveness of IoT and VR for Real–Time Teleoperation of Industrial Robots,” Research Square, 2024

    work page 2024

  26. [26]

    Secure Feder- ated Learning for Cloud–Fog Automation: Vulnerabilities, Challenges, Solutions, and Future Directions,

    Z. Zhang, L. Wu, J. Jin, E. Wang, B. Liu, and Q. Han, “Secure Feder- ated Learning for Cloud–Fog Automation: Vulnerabilities, Challenges, Solutions, and Future Directions,” IEEE Transactions on Industrial Informatics, p. 1–13, 2025

    work page 2025