Lloyd's Register and Orca AI Put AI Navigation to a Proper Test in the Mediterranean

For years, the maritime industry has talked about AI-assisted navigation. This month, Lloyd's Register (LR) actually went out and measured it. LR conducted a structured, five-day live assessment of Orca AI's computer vision navigation platform on a feeder containership sailing from Gioia Tauro, Italy, to Marsaxlokk, Malta — 828 nautical miles through some of the Mediterranean's busiest and most demanding waters, including the Strait of Messina and congested port approaches. This wasn't a vendor demo or a simulation. An LR ship performance specialist joined the vessel as the onboard assessor, collecting real-time feedback on system usability and detection performance while the AI was benchmarked against radar, AIS, and the crew's own visual watchkeeping. The results were strong. The system logged 739 relevant targets across 98 structured observation sessions. It achieved 94% precision — meaning 635 out of 739 objects were correctly detected — and 98.6% recall, picking up nearly everything in the water. There was zero system downtime during the entire voyage. Critically, the AI platform detected small, unlit, and low-radar-signature vessels that traditional radar missed entirely — including non-AIS fishing boats and small craft in night conditions and congested traffic. 63% of observations were conducted in congested waters, where the workload on human watchkeepers is highest and the risk of missing something is greatest. Orca AI's SeaPod units, mounted on top of the bridge, combine day and thermal cameras to provide up to 360-degree coverage, acting as a digital watchkeeper that detects, classifies, and estimates the distance to targets in real time. The display sits centrally in the bridge console. Over 1,200 vessels now use Orca AI systems globally. What makes this trial different is that LR didn't just test detection accuracy — they built a replicable evaluation framework combining precision and recall metrics with structured human factors workshops and crew usability feedback. This framework is designed so that other class societies, shipowners, and technology developers can use the same methodology to assess AI navigation systems on their own vessels. The trial essentially creates a benchmark standard for how the industry should evaluate AI on the bridge. Why this matters for ship managers: If you're waiting for evidence that AI navigation works in real conditions, this is the closest thing to an independent assessment the industry has produced. The LR framework also gives you a way to compare AI navigation products against each other using consistent metrics — something that has been missing from the vendor conversation. For fleet operators already working with Orca AI or evaluating alternatives, the 94% precision and 98.6% recall numbers are your reference point. And the fact that the system caught targets that radar missed in congested Mediterranean waters tells you something important about what your bridge team is currently not seeing. [Lloyd's Register · Splash247 · Hellenic Shipping News]

This story is part of the Maritime AI Digest of 10 May 2026.