Path planning and collision avoidance technologies for maritime autonomous surface ships: a review of COLREGs compliance, algorithmic trends and the navigation-GPT framework
Ho Namgung, J. I. Kim, Da-Un Jang
Abstract
Abstract Safe navigation of maritime autonomous surface ships (MASS) relies on two capabilities: path planning and collision avoidance. This review surveys classical algorithms and modern AI techniques for embedding the International Regulations for Preventing Collisions at Sea (COLREGs) into autonomous navigation. We organise prior work into three families—classical search/optimisation, real-time reactive methods, and learning-based approaches—and discuss their strengths and limitations with respect to rules compliance, computational cost, and onboard constraints. Building on these insights, we outline a large-language-model framework, Navigation-GPT, which couples reasoning-and-acting (ReAct) prompting with low-rank adaptation (LoRA). We further propose a three-phase deployment roadmap for MASS: core model integration, domain fine-tuning, and integrated operations. The paper concludes with open challenges and research directions toward reliable, explainable, and fully compliant MASS navigation.