**Session Date/Time:** 21 Mar 2025 02:30

# nmrg Meeting - IETF 112

## Summary

The NMRG meeting at IETF 112 featured several presentations on Network Digital Twins and the application of AI/ML technologies, particularly Large Language Models (LLMs), for network management and automation. Discussions focused on research challenges, practical considerations, and the role of humans in the management loop. The group also discussed how to best approach the topic of AI/ML within the NMRG context. Laurent announced his departure as co-chair.

## Key Discussion Points

*   **Network Digital Twins:**
    *   Updates on the "Network Digital Twin Concept and Reference Architecture" draft.
    *   Practical considerations and implementation examples of Network Digital Twins, including use cases for SDN, Quantum Key Distribution, and 5G systems.
    *   The difference between simulation and emulation in the context of Network Digital Twins.
    *   Data generation and modeling techniques for building and improving Network Digital Twins.
*   **AI for Network Management:**
    *   General overview of AI for network operation, highlighting opportunities and challenges for IETF/IRTF.
    *   Categorization of use cases in assurance, automation, and optimization.
    *   Discussion on the role of dynamic workflow and agent-based systems in realizing these use cases.
    *   The role of LLMs in network management, focusing on keeping human operators in the loop.
    *   The challenge of semantic gap between network data and AI models.
    *   Security considerations for AI-driven network management.
*   **Research Group Direction:**
    *   The need for a focused research agenda for the group regarding AI/ML, separating standardization-oriented work from research-oriented work.
    *   The importance of identifying specific research questions and challenges that the group can address.

## Decisions and Action Items

*   The "Network Digital Twin Concept and Reference Architecture" draft is ready for IRG review.
*   Reza will look into the draft on the network management agent agent in MOP from Sishing to see about potential collaboration.
*   Offers of the AI/ML presentations should refine and articulate research questions and challenges.
*   Reza needs to determine is it possible to break down the current project into the personalities that Laura just mentioned and see if it's possible.

## Next Steps

*   Authors to refine their work based on meeting feedback.
*   Discussions to continue on the mailing list.
*   Explore potential collaborations between different projects.
*   IRTF chair to see how to make a for a new co-chair for the group.