<ul><li>The O-RAN architecture supports dynamic service demands in wireless networks by utilizing modular, disaggregated components like RAN Intelligent Controller (RIC), Centralized Unit (CU), and Distributed Unit (DU).</li><li>Deep reinforcement learning (DRL) is beneficial for resource allocation and slicing in O-RAN networks but struggles with processing raw input such as RF features and QoS metrics, impacting policy generalization.</li><li>To address these limitations, ORAN-GUIDE introduces a dual-LLM framework that enhances multi-agent RL with structured prompts generated by a domain-specific language model, ORANSight.</li><li>Experimental results demonstrate that ORAN-GUIDE improves sample efficiency, policy convergence, and performance generalization compared to standard MARL and single-LLM approaches.</li></ul>

ORAN-GUIDE: RAG-Driven Prompt Learning for LLM-Augmented Reinforcement Learning in O-RAN Network Slicing

Discover more