Key Takeaways
- An AI‑based computer‑vision system was trialed on a feeder containership sailing through the Mediterranean’s busiest lanes, demonstrating its ability to detect close‑range, low‑signature targets that radar, AIS, and visual watchkeeping sometimes miss.
- The system improved situational awareness during challenging conditions such as night operations, non‑AIS vessel encounters, and small‑craft traffic, supporting watchkeepers in making safer navigation decisions.
- Onboard assessment by LR’s Ship Performance Specialist Han Beng Koe confirmed the technology’s real‑world performance potential and highlighted its scalability for broader maritime adoption.
- Human‑factors workshops led by LR’s Stephanie McLay ensured crew feedback was systematically gathered, analyzed, and applied, underscoring that usability is as critical as detection accuracy for AI‑enabled navigation tools.
- The trial combined precision‑recall metrics with structured human‑factor input, creating a repeatable evaluation framework that can guide shipowners, technology developers, and regulators as AI becomes more embedded in maritime operations.
- Orca AI’s CTO Dor Raviv emphasized that AI‑assisted navigation is already delivering measurable value, citing over 1,200 vessels using the platform as evidence of its safety and operational benefits.
Overview of the Trial
The assessment centered on how AI‑based computer vision can augment human decision‑making in real‑world navigation, especially in complex scenarios like congested waters and reduced visibility. Conducted over a five‑day voyage, the trial took place aboard a feeder containership traveling from the Italian port of Gioia Tauro to Marsaxlokk, Malta—one of the Mediterranean’s busiest shipping corridors. The objective was to evaluate the system’s object‑detection capabilities alongside traditional navigation aids such as radar, AIS, and visual watchkeeping, thereby gauging its practical utility in live operations.
Detection Performance and Operational Benefits
During the voyage, the AI platform consistently identified close‑range and low‑signature targets that were not always visible on conventional sensors. This included non‑AIS vessels, small craft, and objects obscured by darkness or adverse weather. By providing early and accurate detection, the system gave watchkeepers additional reaction time, enhancing safety during night transits and in high‑traffic zones where conventional sensors may suffer from blind spots or signal interference. The improved detection directly supported more informed bridge decisions, reducing the risk of close‑quarters situations.
Onboard Assessment and Human Factors Feedback
LR Ship Performance Specialist Han Beng Koe served as the onboard assessor, offering real‑time observations on the system’s usability and performance. Koe noted that the demonstrated capabilities of AI‑based computer vision within the operational environment clearly indicated its performance potential and scalability for future maritime navigation systems. His feedback bridged the gap between technical metrics and the practical experience of seafarers, validating that the technology functions as intended under genuine sailing conditions.
Human‑Factors Workshops and Usability Research
To complement the technical evaluation, LR facilitated targeted human‑factors workshops aimed at capturing crew insights effectively. Led by Stephanie McLay, Team Lead – Human Factors at LR, the sessions focused on best practices in usability research, ensuring that feedback from seafarers operating under demanding conditions was systematically collected, analyzed, and acted upon. McLay emphasized that the value of AI extends beyond what the technology can do; it hinges on how well it supports the human operator, and structured feedback combined with user‑centered design is essential for creating safer, more usable AI‑enabled navigation tools.
Evaluation Framework: Metrics and Crew Input
The project introduced a structured approach for assessing enhanced situational‑awareness systems, blending objective performance metrics with subjective human‑factor data. Precision and recall were used to quantify detection accuracy, while structured crew feedback addressed usability, trust, and workload impact. This dual‑metric framework aims to provide a holistic view of system effectiveness, offering shipowners, developers, and regulators a reliable reference point for evaluating AI technologies in maritime contexts.
Implications for Maritime Autonomous Shipping
Dor Raviv, CTO and co‑founder of Orca AI, highlighted that the trial demonstrates AI‑assisted navigation is no longer a futuristic concept but a present‑day deliverer of measurable value. With over 1,200 vessels already employing Orca AI, the evidence shows that earlier and more accurate detection leads to better‑informed bridge decisions and safer navigation. Trials such as this one pave the way for broader AI adoption across the industry, supporting the gradual transition toward autonomous shipping while maintaining a strong emphasis on safety and human oversight.
Conclusion and Path Forward
The Mediterranean trial successfully validated AI‑based computer vision as a practical aid for navigators, proving its ability to detect targets missed by traditional sensors and to enhance situational awareness under challenging conditions. By integrating rigorous performance metrics with focused human‑factors research, the study offers a replicable model for evaluating emerging navigation technologies. As decarbonization and autonomy converge in maritime strategy, such evidence‑based assessments will be crucial for guiding investment, regulation, and training efforts, ensuring that AI serves as a trustworthy partner to the human operator on the bridge.