how to exclude stale vessel positions from AI arrival analysis?

Mark and filter AI inputs so AI arrival analysis does not use an old position as the vessel's current location, by excluding stale vessel position records based on position age and data-quality rules. This directly improves data confidence for IT Managers and Fleet Managers by preventing AI arrival models from ingesting outdated AIS position age, stale tracking points, or delayed satellite update records that no longer represent the vessel’s true location.

How Stale Vessel Positions Are Excluded

• Define a position freshness threshold per data source (AIS, radar-derived, satellite) using position timestamp age and store it as a computed field such as “positionAgeMinutes”.
• Exclude records where positionAgeMinutes exceeds the configured limit, and optionally apply a second gate for stale tracking points where movement is inconsistent with the last known track.
• Add data-quality flags at ingestion time (for example, invalid coordinates, missing course and speed, or low confidence) and filter AI training and scoring datasets using those flags.
• Ensure the AI pipeline consumes only the “current position candidate” selected by governance rules, not the most recent record by arrival time, to avoid delayed satellite update artifacts.
• Validate the filter logic with a replay test: run arrival analysis on historical voyages and confirm that excluded points correlate with known gaps or outages in tracking.

Operational Impact

1. Better arrival ETA reliability for Fleet Managers by reducing false “arrival” triggers caused by outdated coordinates, which improves operational planning and reduces unnecessary dispatch changes.
2. Stronger IT governance for AI readiness by standardizing how position age is calculated and enforced across integrations, lowering the risk of inconsistent filtering between systems and reports.
3. Improved auditability and corrective action tracking for data issues by retaining the reason codes for exclusion (for example, “age exceeded threshold” or “invalid coordinates”), enabling targeted troubleshooting and data stewardship.

Important to know: Treat “position age” as a first-class data element in your master data and event model, and align the freshness threshold with your operational tracking expectations (for example, expected AIS update cadence and typical satellite reporting delays). Then log every exclusion decision so IT and Fleet teams can trace why a vessel’s position was not used in AI arrival analysis.