VerifAI: framework for functional verification of AI-based systems in the maritime domain

• Autorzy: Stach Thomas , Koch Paul , Constapel Manfred , Portier Martin , Schmid Helmut

Identyfikatory

DOI

10.12716/1001.18.03.12

• Języki publikacji: EN

Abstrakty

EN

With the continuous emergence and steady development of new technologies the way for Maritime Autonomous Surface Ship (MASS) is being paved. However, this manifold of available and imminent technologies challenges regulatory bodies and auditing authorities. Technologies which make use of Artificial Intelligence (AI), in particular Machine Learning (ML), play a special role. On one hand, they are not covered by current regulations or audit processes and, on the other hand, they may represent black boxes whose behaviours are not readily explainable and thus impede audit processes even further. In an upcoming study titled VerifAI the authors focus on this gap within European and German regulatory bodies and auditing authorities. The technological scope lies on MASS-related products which rely on partially or fully AI-based systems. In the present article the original authors summarize the outlined study. The authors review the current regulatory status concerning audit processes and the market situation concerning available and imminent (partially) AI-based systems of MASS-related products. To close the gap a conceptual, integrated framework consisting of a Safety Guideline for the manufacturers and a Verification Guideline for the auditing authorities is presented. The framework aims to give regulatory bodies and auditing authorities an overview of necessary steps for robust verification of safe products without hindering innovation or requiring in-depth knowledge about the (black box-like) systems. The results are condensed into recommendations for actions, listing the most important results, and proposing entry points as well as future research in the field of verifying (partially) AI-based MASS-related products.

Słowa kluczowe

EN

methodological framework; Safety of Life at Sea; SOLAS; SOLAS Convention; maritime domain; Maritime Autonomous Surface Ships; MASS; Artificial Intelligence; AI; Machine Learning; AI-based systems; Artificial Intelligence Act

• Wydawca: Faculty of Navigation, Gdynia Maritime University
• Czasopismo: TransNav : International Journal on Marine Navigation and Safety of Sea Transportation
• Rocznik: 2024
• Tom: Vol. 18 No. 3
• Strony: 585--591
• Opis fizyczny: Bibliogr. 25 poz., rys.

Twórcy

autor

Stach Thomas

• Fraunhofer Centre for Maritime Logistics and Services CML, Hamburg, Germany

• https://orcid.org/0009-0008-9920-1297

autor

Koch Paul

• Fraunhofer Centre for Maritime Logistics and Services CML, Hamburg, Germany

• https://orcid.org/0000-0001-7181-8529

autor

Constapel Manfred

• Fraunhofer Centre for Maritime Logistics and Services CML, Hamburg, Germany

• https://orcid.org/0000-0002-7015-6778

autor

Portier Martin

• Federal Maritime and Hydrographic Agency of Germany, Hamburg, Germany

autor

Schmid Helmut

• Federal Maritime and Hydrographic Agency of Germany, Hamburg, Germany

Bibliografia

• [1] S. K. Brooks and N. Greenberg, “Mental Health and Psychological Wellbeing of Maritime Personnel: A Systematic Review,” BMC Psychology, vol. 10, no. 1, pp. 1–26, 2022.
• [2] C. Berghoff, B. Biggio, E. Brummel, V. Danos, T. Doms, H. Ehrich, T. Gantevoort, B. Hammer, J. Iden, S. Jacob, H. Khlaaf, L. Komrowski, R. Kröwing, J. H. Metzen, M. Neu, F. Petsch, M. Poretschkin, W. Samek, H. Schäbe, A. V. Twickel, M. Vechev, T. Wiegand, W. Samek, and M. Fliehe, “Towards Auditable AI Systems,” Whitepaper, 2021.
• [3] W. Samek and K.-R. Müller, “Towards Explainable Artificial Intelligence,” in Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2019, vol. 11700 LNCS, pp. 5–22.
• [4] Europäisches Parlament und Rat der Europäischen Union, “Richtlinie 2014/90/EU des europäischen Parlaments und des Rates vom 23. Juli 2014 über Schiffsausrüstung und zur Aufhebung der Richtlinie 96/98/EG des Rates (2014/90/EU),” pp. 146–185, 2014.
• [5] E. Kommission, “Vorschlag für eine Verordnung des Europäischen Parlaments und des Rates zur Festlegung harmonisierter Vorschriften für künstliche Intelligenz (Gesetz über künstliche Intelligenz) und zur Änderung bestimmter Rechtsakte der Union,” 2021.
• [6] B. Rokseth, O. I. Haugen, and I. B. Utne, “Safety Verification for Autonomous Ships,” MATEC Web of Conferences, vol. 273, 2019.
• [7] H. Ringbom, “Regulating Autonomous Ships—Concepts, Challenges and Precedents,” Ocean Development & International Law, vol. 50, no. 2-3, pp. 141–169, 2019.
• [8] IMO, “Maritime safety committee (MSC 105),” https://www.imo.org/en/MediaCentre/MeetingSummaries/Pages/MSC-105th-session.aspx, 2022.
• [9] International Maritime Organization, “International Convention for the Safety of Life at Sea,” 1974.
• [10] “Google Patents on (’Autonomous’ AND ’Ship’),” https://patents.google.com/.
• [11] IMO, “Resolution MSC.192(79), Adoption of the Revised Performance Standards for Radar Equipment,” International Maritime Organization, Tech. Rep., 2004.
• [12] IMO, “Resolution A.1106(29), Revised Guidelines for the Onboard Operational Use of Shipborne Automatic Identification Systems (AIS),” International Maritime Organization, Tech. Rep., 2015.
• [13] H.-C. Burmeister, M. Constapel, C. Uge´, and C. Jahn, “From Sensors to MASS: Digital Representation of the Perceived Environment Enabling Ship Navigation,” ser. IOP Conference Series: Materials Science and Engineering, vol. 929. IOP Publishing, 2020.
• [14] M. Gyllenhammar, R. Johansson, F. Warg, D. Chen, H.-M. Heyn, M. Sanfridson, J. Söderberg, A. Thorsen, S. Ursing, Z. Ab, and M. G. Com, “Towards an Operational Design Domain that Supports the Safety Argumentation of an Automated Driving System,” 10th European Congress on Embedded Real Time Systems, pp. 1–10, 2020.
• [15] Ø. J. Rødseth, L. A. L. Wennersberg, and H. Nordahl, “Towards Approval of Autonomous Ship Systems by Their Operational Envelope,” Journal of Marine Science and Technology, vol. 27, no. 1, pp. S. 67–76, 2022.
• [16] DIN, “DIN EN 61162-1:2011-09 Navigations- und Funkkommunikationsgeräte und -systeme für die Seeschifffahrt - Digitale Schnittstellen - Teil 1: Ein Datensender und mehrere Datenempfänger,” DIN Deutsches Institut für Normung e. V., Tech. Rep., 2011.
• [17] M. Korakakis, P. Mylonas, and E. Spyrou, “A Short Survey on Modern Virtual Environments That Utilize AI and Synthetic Data,” ser. MCIS 2018 Proceedings, 2018, p. 34.
• [18] S. I. Nikolenko, Synthetic Data for Deep Learning. Springer International Publishing, 2021, vol. 174.
• [19] A. Tsirikoglou, J. Kronander, M. Wrenninge, and J. Unger, “Procedural Modeling and Physically Based Rendering for Synthetic Data Generation in Automotive Applications,” 2017.
• [20] A. Ramesh, P. Dhariwal, A. Nichol, C. Chu, and M. Chen, “Hierarchical text-conditional image generation with clip latents,” ArXiv, vol. abs/2204.06125, 2022.
• [21] R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High-resolution image synthesis with latent diffusion models,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022.
• [22] M. Z. Naser and A. H. Alavi, “Error Metrics and Performance Fitness Indicators for Artificial Intelligence and Machine Learning in Engineering and Sciences,” Architecture, Structures and Construction, 2021.
• [23] V. N. Gudivada, J. Ding, and A. Apon, “Data Quality Considerations for Big Data and Machine Learning: Going Beyond Data Cleaning and Transformations,” International Journal on Advances in Software, vol. 10.1, pp. 1–20, 2017.
• [24] A. Navlani, A. Fandango, and I. Idris, Python Data Analysis: Perform Data Collection, Data Processing, Wrangling, Visualization, and Model Building Using Python, third edition ed. Birmingham: Packt Publishing, 2021.
• [25] DIN, “DIN EN ISO/IEC 23053 Framework for Artificial Intelligence (AI) Systems Using Machine Learning (ML),” DIN Deutsches Institut für Normung e. V., Tech. Rep., 2023.

Uwagi

1. Pełne imiona podano na stronie internetowej czasopisma w "Authors in other databases."

2. Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki (2025).