Multi-Agent Digital Twin for SLA Management in Coherent P2MP Metro-Access Networks

Dynamic multi-service traffic in coherent point- to-multipoint (P2MP) metro-access networks makes proactive service level agreement (SLA) management increasingly challeng- ing. Existing AI-enabled network management solutions primar- ily support reactive monitoring and provide limited capability for anticipating future SLA degradation. This demonstration presents an AI-native digital twin for predictive SLA manage- ment that integrates traffic forecasting, SLA-state prediction, recovery recommendation, and a large language model (LLM)- assisted conversational interface into a unified decision-making framework. By predicting future enterprise, radio access network (RAN), and passive optical network (PON) traffic demands, the framework identifies potential SLA violations, estimates network risk, and recommends proactive mitigation actions before service degradation occurs. Through natural-language interactions, oper- ators can query future network conditions, understand the causes of predicted SLA violations, and receive explainable operational recommendations. The demonstration highlights the potential of AI-native digital twins to enable proactive, intent-driven, and explainable management of next-generation coherent P2MP metro-access networks.