Integrate an agent for each element into a closed-loop, coherent AFDDR architecture, and build a modular simulation framework to evaluate it quantitatively.
How do the interactions between agents achieve successful anomaly response — including which recovery procedure to try next?
How much does the system reduce the need for human intervention, and how far does it achieve autonomy?
How do alternative diagnostic methods compare inside the same loop?