Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Harnessing the Power of LLMs, Informers and Decision Transformers for Intent-driven RAN Management in 6G

Md Arafat Habib, Pedro Enrique Iturria Rivera, Yigit Ozcan, Medhat Elsayed, Majid Bavand, Raimundas Gaigalas, Melike Erol-Kantarci

TL;DR

This paper tackles the complexity of intent-driven RAN management in 6G by proposing a three-step GenAI framework that translates operator intents into network actions. It leverages a memory-efficient QLoRA-tuned LLM for intent processing, an Informer-based predictive module for proactive intent validation, and the Hierarchical Decision Transformer with Goal Awareness (HDTGA) for orchestrating network applications. The approach achieves measurable improvements: a 6% boost in BERTScore, a 9% rise in semantic similarity, 88% accuracy in predictive validation, and substantial gains in throughput (≥19.3%), delay reduction (≥48.5%), and energy efficiency (≥54.9%) over baselines. This highlights the framework’s potential to enable adaptive, on-demand service deployment in dynamic, resource-intensive 6G environments.

Abstract

Intent-driven network management is critical for managing the complexity of 5G and 6G networks. It enables adaptive, on-demand management of the network based on the objectives of the network operators. In this paper, we propose an innovative three-step framework for intent-driven network management based on Generative AI (GenAI) algorithms. First, we fine-tune a Large Language Model (LLM) on a custom dataset using a Quantized Low-Rank Adapter (QLoRA) to enable memory-efficient intent processing within limited computational resources. A Retrieval Augmented Generation (RAG) module is included to support dynamic decision-making. Second, we utilize a transformer architecture for time series forecasting to predict key parameters, such as power consumption, traffic load, and packet drop rate, to facilitate intent validation proactively. Lastly, we introduce a Hierarchical Decision Transformer with Goal Awareness (HDTGA) to optimize the selection and orchestration of network applications and hence, optimize the network. Our intent guidance and processing approach improves BERTScore by 6% and the semantic similarity score by 9% compared to the base LLM model. Again, the proposed predictive intent validation approach can successfully rule out the performance-degrading intents with an average of 88% accuracy. Finally, compared to the baselines, the proposed HDTGA algorithm increases throughput at least by 19.3%, reduces delay by 48.5%, and boosts energy efficiency by 54.9%.

Harnessing the Power of LLMs, Informers and Decision Transformers for Intent-driven RAN Management in 6G

TL;DR

This paper tackles the complexity of intent-driven RAN management in 6G by proposing a three-step GenAI framework that translates operator intents into network actions. It leverages a memory-efficient QLoRA-tuned LLM for intent processing, an Informer-based predictive module for proactive intent validation, and the Hierarchical Decision Transformer with Goal Awareness (HDTGA) for orchestrating network applications. The approach achieves measurable improvements: a 6% boost in BERTScore, a 9% rise in semantic similarity, 88% accuracy in predictive validation, and substantial gains in throughput (≥19.3%), delay reduction (≥48.5%), and energy efficiency (≥54.9%) over baselines. This highlights the framework’s potential to enable adaptive, on-demand service deployment in dynamic, resource-intensive 6G environments.

Abstract

Intent-driven network management is critical for managing the complexity of 5G and 6G networks. It enables adaptive, on-demand management of the network based on the objectives of the network operators. In this paper, we propose an innovative three-step framework for intent-driven network management based on Generative AI (GenAI) algorithms. First, we fine-tune a Large Language Model (LLM) on a custom dataset using a Quantized Low-Rank Adapter (QLoRA) to enable memory-efficient intent processing within limited computational resources. A Retrieval Augmented Generation (RAG) module is included to support dynamic decision-making. Second, we utilize a transformer architecture for time series forecasting to predict key parameters, such as power consumption, traffic load, and packet drop rate, to facilitate intent validation proactively. Lastly, we introduce a Hierarchical Decision Transformer with Goal Awareness (HDTGA) to optimize the selection and orchestration of network applications and hence, optimize the network. Our intent guidance and processing approach improves BERTScore by 6% and the semantic similarity score by 9% compared to the base LLM model. Again, the proposed predictive intent validation approach can successfully rule out the performance-degrading intents with an average of 88% accuracy. Finally, compared to the baselines, the proposed HDTGA algorithm increases throughput at least by 19.3%, reduces delay by 48.5%, and boosts energy efficiency by 54.9%.
Paper Structure (25 sections, 1 theorem, 37 equations, 12 figures, 3 tables, 4 algorithms)

This paper contains 25 sections, 1 theorem, 37 equations, 12 figures, 3 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Intent Processing
  4. Intent Validation
  5. Intent Execution via Application Orchestration
  6. System Model and Problem Formulation
  7. System Model
  8. Network Elements
  9. Controller Architecture
  10. Applications
  11. Problem formulation
  12. Intent and Query Processing by
  13. Intent Validation
  14. Application Orchestration
  15. Proposed Three-Step Methodology for Intent-driven Network Management
  16. ...and 10 more sections

Key Result

Theorem 1

The Hierarchical Decision Transformer with Goal Awareness (HDTGA) learns a policy $\pi_{\theta}(a_t | s_{\leq t}, g_t)$ such that the expected return $V^{\pi_{\theta}}(s_0, g_0)$ is within $\epsilon$ of the optimal expected return $V^{*}(s_0, g_0)$ for all initial states $s_0$ and goals $g_0$, i.e., provided that the model capacity and training procedure are sufficient to minimize the empirical lo

Figures (12)

  • Figure 1: Network and system model.
  • Figure 2: Step-by-step illustration of the proposed methodology.
  • Figure 3: Example queries/intents with responses.
  • Figure 4: Fine-tuned LLM model with RAG.
  • Figure 5: HDTGA architecture with an offline dataset collected from the network environment.
  • ...and 7 more figures

Theorems & Definitions (1)

  • Theorem 1