Network Function Orchestration with LLM based Multi-Agent System

Large Language Models (LLMs) have significantly advanced the field of Natural Language Processing (NLP), enabling breakthroughs in contextual understanding, language generation, and reasoning. LLMs possess an intrinsic ability to act as orchestrators in dynamic and multifaceted environments. This work introduces a novel framework for dynamic function orchestration that leverages the reasoning capabilities of LLMs through Chain-of-Thought (CoT). The proposed framework comprises a sequence of three agents: a Semantic Router to direct user requests to the corresponding administrative domain, an Intent Translator to generate strategic plans utilizing a knowledge base containing 3GPP specifications, and a Resolver that selects and executes functions to fulfill the strategic plan. We systematically analyze the strengths and limitations of LLMs in collaborative orchestration, and demonstrate their capability to dynamically adapt to complex decision-making scenarios in the Telecom domain.