Agentic6G · AUTONOMOUS MULTI-AGENT AGENTIC AI SYSTEM FOR 6G NETWORKS

Operating over B5G/6G networks, the Agentic6G multi-agent Agentic-AI system (MAS) will be able to benefit from and further evolve the interplay of 6G network and compute across the Device-Edge-Cloud continuum and potentially from new 6G architectural advancements such as the AI-native paradigm, in addition to the 6G extreme communication capabilities beyond 5G. For service composition, Agentic6G will allow the MAS to autonomously compose, coordinate, and optimise B5G/6G and/or vertical business microservices across Edge-Cloud infrastructures, advancing the B5G cloud-native approach for autonomous service composition, deployment and management, by combining collective agentic AI’s knowledge handling as an integral element through fast, efficient, and autonomous access to and application of knowledge with the support of Large Language Models (LLMs), shared memories and tools. Agentic6G will enable efficient end-to-end Agentic Operations (AgenticOps) to support the lifecycle management of MAS to enable this vision. Furthermore, to meet the challenging and demanding operational environments in real world, Agentic6G will provide self-organising capabilities focusing on spontaneous agent generation, and self-protection, self-healing and auto-scaling of multi-agent systems through hybrid orchestration featured with agent-embedded orchestration for decentralised scenarios and advanced self-organising capabilities as the agents’ tools. Thereby, Agentic6G will help remove the barriers for adopting and deploying such multi-agent agentic AI systems in B5G and 6G networks, by addressing some of the well-known and critical challenges such as security risks, and lack of availability/robustness/resilience and scalability. The main outcomes will be a set of Agentic 6G agent as autonomy enablers and an integrated Agentic6G MAS, tested, validated and demonstrated with representative technical use cases and a complicated swarm intelligent robotics vertical application.