Manrova: What We Learned Building a Multi Agent System That Watches, Doesn't Just Answers

Описание и применение

Manrova – система многопользовательских агентов, непрерывно наблюдающая за морскими судами, выявляющая аномалии и передающая решения человеку только при необходимости.

Задача

Vessel operators lack time to correlate GPS, radar, crew schedules, near‑miss reports, and compliance records, leading to incidents instead of warnings.

Результат

Manrova achieved continuous, deterministic monitoring of vessel operations, reducing incident reports by enabling early warnings and human‑in‑the‑loop decision making.

Рабочий процесс

• 1. Navigation Integrity Agent checks GPS vs radar positions.
• 2. Crew Readiness Agent monitors rest/duty patterns for fatigue risk.
• 3. Fleet Pattern Agent scans fleet‑wide near‑miss history.
• 4. Compliance Readiness Agent evaluates certificate exposure relative to schedules.
• 5. Officer of the Watch (OOW) orchestrates agents, correlates evidence, applies deterministic risk logic.
• 6. If severity below threshold, system continues silent monitoring; if above, OOW prepares response and awaits human approval.

Доказательства

Safety‑critical math never touches a language model. Position deviation, weighted risk scoring, incident state transitions: all deterministic Python, computed the same way every time, auditable, reproducible.

Источник

The hard part was deciding, agent by agent, exactly what each one was allowed to trust, when it should hand off to another agent, and where control had to return to a human no matter what the model said.

Источник