Huawei's MoM-based Multi-Agent Collaboration System Empowers Core Networks' Transitioning to L4 High Stability

SHENZHEN, China, Dec. 22, 2025 /PRNewswire/ -- In an increasingly digital era, modern life and economic activities heavily rely on networks. Network stability directly impacts public 'digital well-being' and 'sense of social participation.' As the brain of the network, the core network must remain stable, since any service interruption can affect tens of millions of users, causing immeasurable societal impact. Huawei's ICNMaster MDAF solution builds an intelligent high-stability system based on MoM (Mixture of Models), multi-agent collaboration, and network digital twin technologies. It enables proactive risk prevention, rapid fault recovery, and service restoration within minutes, helping global operators accelerate their journey toward L4 high-stability and ensuring service continuity experience.

As 5G and cloud technologies evolve, services become more diversified. Alongside their emergence come software-hardware decoupling, multi-generation coexistence, complex inter-generation APIs, signaling surges, transport network faults, and data center failures, all of which exert greater pressure on operators to maintain carrier-grade reliability of the core network. According to GlobalData statistics, 42% of operators have experienced core network service interruptions in the past three years, with the frequency increasing annually, drawing widespread industry concern. To guide global operators in progressively improving core network reliability, TM Forum has released a core network high-stability assessment standard. Dozens of operators have already participated in evaluations. Core network reliability enhancement has become an industry consensus.

To meet global operators' demands for core network high stability, Huawei launched the ICNMaster MDAF intelligent high-stability system. Building upon earlier innovations like the "Fault Management Agent" and "Complaint Handling Agent," it introduces breakthrough technologies: the MoM architecture, multi-agent collaboration, and agent-network digital twin synergy. This enables automated problem closure, reduces incident probability, achieves service recovery in minutes, and empowers core network L4 high stability.

From 'Single-Model Architecture' to 'Multi-Model Architecture': Compared to traditional single-model architectures, the MoM architecture achieves scenario breakthroughs. It fully integrates the advantages of the fast inference model with the deep reasoning model like DeepSeek. An intelligent model traffic routing framework dynamically allocates tasks: routine, high-frequency events are handled instantly by fast, accurate models, while complex anomalies requiring deep reasoning are assigned to reasoning models. This ensures lightning-fast responses to common issues and deep reasoning for complex faults, achieving optimal efficiency and precision.

From 'Single-Agent Automation' to 'Multi-Agent Collaborative Self-Closure': Traditional single agents only automate isolated individual scenarios. Multi-agent collaboration technology can orderly unite multiple agents, performing agent orchestration, conflict resolution, and more. Recovery actions are validated through the network digital twin system, ensuring corrective measures are accurate and effective before being applied to the live network, ultimately achieving a self-closing loop of "perception-analysis-decision-execution."

Huawei's ICNMaster MDAF MoM-based multi-agent collaborative high-stability solution goes beyond an enhancement; it is a crucial step toward fully autonomous networks. It enables operators to provide not just connectivity, but a future-oriented, resilient, and intelligent digital foundation. It successfully resolves the long-standing conflict between network stability and operational efficiency and marks a fundamental strategic shift in the telecom industry's operational model from network-centric to user-centric. The 'Agent + Digital Twin' collaboration paradigm validated by this system lays a solid technical and practical foundation for realizing fully autonomous driving networks in the future.