Implementation and Application of a Knowledge Service for AUV Mission Explainability

Abstract

This paper presents a knowledge service aimed at enhancing mission explainability in subsea robotics operations. The proposed system consists of an AI agent that chains together specialized large language models (LLMs) and a graph database to enable natural language querying and interactive visualization. The graph database models entities and relationships relevant to subsea inspection and maintenance, such as clients, industries, sites, vehicles, sensors, and underwater scene elements, including pipeline components and seafloor characteristics. The browser-based GUI aims to allow stakeholders—including field teams, robot developers, industry clients, and regulatory agencies—to intuitively interact with mission data, supporting post-mission analysis and explainability. Using pipeline inspections as a case study, we illustrate the potential of this approach and discuss future developments needed to advance this framework toward a practical solution for subsea robotics.