Trustworthy AI-Driven Autonomous Underwater Vehicles for Port Infrastructure Inspection: Paradigm Conceptualization

This paper conceptualizes a Trustworthy AI paradigm for Autonomous Underwater Vehicles (AUVs) tasked with port-infrastructure inspection and maintenance, instantiated through the FAITH (Fostering Artificial Intelligence Trust for Humans) Trustworthiness Assessment Framework (AI_TAF)-an operational playbook aligning technical, legal, and societal criteria. Within our Large-Scale Pilot (LSP), we present Automatic Target Recognition (ATR) for detection and geolocation of objects/anomalies and Autonomous Coverage (AC) for adaptive multibeam-based bathymetric surveying. We map core trustworthiness attributes for both AI systems-validity/reliability, safety, explainability/interpretability, fairness, and accountabil-ity/transparency-into verifiable requirements, then conduct AI asset cartography and a threat-consequence-vulnerability analysis to derive risk controls.