Comparison Of AIS-Based Prediction Of The Distance At The CPA With Factual Separation Between Vessels

Banyś, Paweł; Heymann, Frank; Engler, Evelin; Noack, Thoralf · 2014 · Crossref

DOI: 10.1515/aon-2015-0001

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study evaluates the reliability of Automatic Identification System (AIS) data for predicting the Closest Point of Approach (CPA) distance and Time to CPA (TCPA) between vessels. Although AIS is widely used for watchkeeping, it is not currently recognized under International Regulations for Preventing Collisions at Sea (Colregs) as an approved anticollision instrument due to concerns regarding data integrity and incomplete transmissions. The research aims to determine if AIS data, in its current form, can accurately appraise traffic situations and support collision avoidance maneuvers by comparing predicted values with factual separation distances. The analysis utilized AIS data collected in September 2011 from the Strait of Fehmarn-Belt, a busy shipping route connecting the Baltic Sea with the rest of the world. The dataset, provided by the German Federal Waterway Authority, contained approximately 99.9 million dynamic AIS messages. After filtering out 0.6% of records with missing parameters, the researchers processed the remaining data using a transverse Mercator projection to simplify distance calculations. To identify relevant vessel encounters, a Delaunay triangulation method linked vessels within 6 nautical miles of each other. This process identified 8,658 distinct encounters, categorized into overtaking, head-on, and crossing situations. For each encounter, the study tracked the evolution of predicted CPA and TCPA values over time and compared them against the actual minimal distances recorded by AIS. The results indicate that AIS-based predictions are viable for collision avoidance if appropriate safety margins are applied. Factual CPA distances ranged from zero to six nautical miles, with peaks at 1 NM (parallel courses) and 0.3 NM (ferry crossings). The study found that CPA distance predictions stabilize close to factual values approximately 20 minutes before the encounter. Specifically, 70% of CPA distance predictions deviated no more than 0.5 NM from the factual distance 20 minutes prior to the encounter, rising to 90% accuracy 10 minutes prior. In contrast, TCPA predictions were more volatile; only 30% of timestamp predictions were within one minute of the actual CPA time 20 minutes prior, increasing to nearly 80% just 10 minutes before the encounter. The analysis showed that relaxing the safety margin for CPA distance significantly improves prediction reliability more effectively than relaxing the margin for TCPA. The authors conclude that AIS data can practically support the prediction of CPA distance and TCPA, provided seafarers account for inherent data noise and potential transmission errors. While CPA distance forecasts are sufficiently reliable for decision-making, TCPA predictions remain less stable until shortly before the encounter. The study suggests that because accurate distance forecasting is more critical for collision avoidance than time forecasting, AIS could be authorized for anticollision tasks if data integrity parameters are improved and transponder performance is enhanced to reduce faulty transmissions.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• naturalistic crash near crash