Condense: Cognitive intent-driven end-to-end network slicing with AI planning agents

Autonomous network management has gained momentum as a key feature of 5G advanced and 6G systems. Autonomy builds on the use of intent-driven networks that involve intelligent agents to configure various network sub-domains. In addition, the concept of end-to-end network slicing would need to be appropriately managed to provide differentiated Quality of Service (QoS) within 5G/6G networks. However, there is limited work looking at intent-driven end-to-end network slice assurance. In this paper, we propose Condense, a cognitive intent-driven system for network slice fulfilment and assurance. Via the use of Artificial Intelligence (AI) planning agents, we demonstrate the decomposition of high-level intent requirements to individual sub-intents. AI planning makes use of machine reasoning techniques to search for issue resolutions that differs from other data-driven approaches. AI planning agents are employed for resource allocation at the Radio Access Network (RAN), transport and core domains to ensure intent fulfilment. The system is implemented over a cognitive intent management system to demonstrate end-to-end network slicing.