Verification and Validation of Autonomous Collision Avoidance System Developed based on Functional Task Analysis of Ship Operation by Ship Maneuvering Experts

The verification and validation (V&V) of autonomous ship systems are crucial to ensure not only their safety, but also that their behavior aligns with the expectations of human operators. However, ensuring such human-like behavior remains a central challenge for autonomous ship systems. This research, to address the issue, aims to develop an autonomous collision avoidance algorithm designed based on expert task analysis. Using the Functional Resonance Analysis Method (FRAM), we developed an algorithm that mimics the flexible and strategic behaviors of human experts. The algorithm was evaluated through quantitative analysis and subjective reviews by experienced captains. Their feedback was used to iteratively refine the algorithm, improving its ability to exhibit human-like and context-sensitive behavior. This paper describes how expert feedback was used to inform the iterative improvement of the algorithm and discusses the insights gained through this human-centered validation process.