A fuzzy-based hybrid model proposal for passage planning cross track limit determination

One of the important safety parameters that must be determined for safe navigation in voyage planning on ships is the Cross Track Limit (XTL) value. These values, determined by considering obstacles on the water, define the safe navigation area on the starboard and port sides of the ship's planned route. In practice, these values are typically set by company policies and/or the ship master. However, personal experience and risk perception significantly influence the established limits. Given that most ship accidents are attributed to human error, accurately determining the XTL value is crucial for ensuring safe navigation. This study aims to determine XTL values using machine learning methods. Initially, factors influencing XTL values were identified through academic studies, company policies, and expert opinions. These factors were categorized under two main headings: open sea and restricted waters. A data collection form, designed based on the obtained information, was presented to 11 oceangoings masters. Using the collected data, the factors affecting the XTL value were ranked in order of importance through the Fuzzy Extended Analytical Hierarchy Process (FE-AHP), a Multi-Criteria Decisions Making (MCDM) method. Then, the five factors with the highest impact were determined. While the first five factors in open seas were determined as ship maneuverability, ship length, ship breadth, watchkeeping officer experience and ship type, in confined waters they were determined as ship draft, CATZOC, traffic situation, watchkeeping officer experience and ship speed. Based on these determined factors, a Mamdani-type Fuzzy Inference System (FIS) model was created. This FIS model was designed to recommend XTL values using machine learning techniques. As a result, XTL values, traditionally determined by ship masters and/or company policies, are recommended by a decision support model based on machine learning, eliminating subjective judgments.