Evaluation of an Assurance Framework for Naval Electro-Optic-based Machine Learning Applications
• Alignment of naval EO assurance with AMLAS to enable trusted ML-based autonomy.
• Progressive assurance tightening across autonomy levels, reflecting the shift from human-in-the-loop to human-on-the-loop and beyond.
• Embedded safety mechanisms: health monitoring, watchdogs, and fault-tolerant operation.
• Application of STPA and SOTIF, to identify hazards from both system malfunctions as well as deployment in unsuitable operating environments, respectively
