Large Language Models Enable Automated Configuration and Resource Optimization in Radio Access Networks

Large Language Models Enable Automated Configuration and Resource Optimization in Radio Access Networks

Researchers have developed frameworks using large language models (LLMs) to automate critical network management tasks in radio access networks (RANs), according to multiple papers published on arxiv.org.

According to one arxiv.org paper, an LLM-assisted framework for automatic gNB parameter configuration addresses a common problem: misconfigurations that cause system failures and traditionally require manual log analysis. The framework uses synthetic data generation following a “configuration, log, correction workflow” and was validated on an OpenAirInterface (OAI) gNB testbed with 480 misconfiguration scenarios. Fine-tuning improved correction accuracy from 13.8% (zero-shot baseline) to 85.4%, with retrieval-augmented generation (RAG) further increasing accuracy to 92.7%.

A separate arxiv.org paper proposes an “Optimization-as-a-Service” approach using multi-agent LLM systems for physical resource block (PRB) allocation in RANs. According to the researchers, this framework addresses challenges in sixth-generation (6G) RAN environments, which face “unprecedented service diversity and exponential dynamics, featuring volatile fluctuations in active base stations (BSs), user scale, and stringent Quality-of-Service (QoS) requirements.” The closed-loop architecture integrates scene understanding, objective generation, solver, and reflection agents, with a one-shot reflection distillation mechanism to reduce computational latency.

Both frameworks demonstrate LLMs’ potential to enable automated, scalable network operations without manual intervention, according to the source materials.