Retrieval-Augmented Generation as an Area-of-Measurable-Performance (AOMP) to Enhance the Effectiveness, Modularity, and Trust of Flight Test Safety Planning
Name
Fuller_Final[47].pdf
Size
2.95 MB
Format
Adobe PDF
Checksum (MD5)
c01d20683edf42da0993daf98a23aabc
Author(s)
Fuller, Samuel
Picardo, Christopher
Johnson, Kevin
James, Michael
Date Issued
August 20, 2026
Abstract
State of the art Test Hazard Analysis (THA) development remains a largely manual and knowledge-intensive process. The current approach relies heavily on subject matter expertise and institutional knowledge specific to previous test programs and to specific aircraft. Recent advances in large language models (LLMs) have made it possible to meaningfully assist in documentation-heavy workflows. However, the concern of adoption of ad hoc non-human in-the-loop Artificial Intelligence (AI) in safety-critical environments risks the introduction of unverifiable outputs, limited traceability, and unclear performance benefits remains. To address these concerns, this paper characterizes THA as an Area of Measurable Performance (AOMP) to implement and evaluate the THA-RAG model, a modular, human-in-the-loop retrieval-augmented generation (RAG) capability against explicitly defined modular performance requirements and metrics to promote the reuse and the level of trust of a wider range of THA applications within the JSE ecosystems of technologies and broader TEST community.
Subjects
Area of Measurable Performance
Artificial Intelligence
Air Force Test Center
Defense Modeling and Simulation
Flight Test Safety Planning
Generally Accepted Accounting Principles
Joint Simulation Environment
Human In The Loop
Large Language Models
Machine Learning
Metrics
Modular Components
Retrieval-Augmented Generation
Requirements
Test
Test Hazard Analysis
Training
Validation
Verification
MIT Department
Lincoln Laboratory
Terms of Use
All rights reserved
Persistent DSpace Link
