Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An LLM-Agent-Based Framework for Age of Information Optimization in Heterogeneous Random Access Networks

Fang Liu, Erchao Zhu, Jiedan Tan, Jingwen Tong, Taotao Wang, Shengli Zhang

TL;DR

This work addresses the challenge of maintaining information freshness in heterogeneous random-access networks by introducing Reflex‑Core, an LLM‑based agent framework with an Observe‑Reflect‑Decide‑Execute loop. It integrates supervised fine‑tuning and PPO to align semantic reasoning with AoI objectives and develops two protocols, the Reflexive Multiple Access (RMA) and a Priority‑Aware RMA, to enable autonomous, differentiated access control. Empirical results show that RMA achieves up to 14.9% reductions in system AoI and 20% faster convergence relative to baselines across diverse scenarios, with the priority variant offering faster convergence and stronger QoS guarantees. The framework demonstrates robustness to dynamic conditions and offers a path toward self‑evolving wireless networks that can adapt to protocol heterogeneity and traffic priorities in real time.

Abstract

With the rapid expansion of the Internet of Things (IoT) and heterogeneous wireless networks, the Age of Information (AoI) has emerged as a critical metric for evaluating the performance of real-time and personalized systems. While AoI-based random access is essential for next-generation applications such as the low-altitude economy and indoor service robots, existing strategies, ranging from rule-based protocols to learning-based methods, face critical challenges, including idealized model assumptions, slow convergence, and poor generalization. In this article, we propose Reflex-Core, a novel Large Language Model (LLM) agent-based framework for AoI-driven random access in heterogeneous networks. By devising an "Observe-Reflect-Decide-Execute" closed-loop mechanism, this framework integrates Supervised Fine-Tuning (SFT) and Proximal Policy Optimization (PPO) to enable optimal, autonomous access control. Based on the Reflex-Core framework, we develop a Reflexive Multiple Access (RMA) protocol and a priority-based RMA variant for intelligent access control under different heterogeneous network settings. Experimental results demonstrate that in the investigated scenarios, the RMA protocol achieves up to a 14.9% reduction in average AoI compared with existing baselines, while the priority-based version improves the convergence rate by approximately 20%.

An LLM-Agent-Based Framework for Age of Information Optimization in Heterogeneous Random Access Networks

TL;DR

This work addresses the challenge of maintaining information freshness in heterogeneous random-access networks by introducing Reflex‑Core, an LLM‑based agent framework with an Observe‑Reflect‑Decide‑Execute loop. It integrates supervised fine‑tuning and PPO to align semantic reasoning with AoI objectives and develops two protocols, the Reflexive Multiple Access (RMA) and a Priority‑Aware RMA, to enable autonomous, differentiated access control. Empirical results show that RMA achieves up to 14.9% reductions in system AoI and 20% faster convergence relative to baselines across diverse scenarios, with the priority variant offering faster convergence and stronger QoS guarantees. The framework demonstrates robustness to dynamic conditions and offers a path toward self‑evolving wireless networks that can adapt to protocol heterogeneity and traffic priorities in real time.

Abstract

With the rapid expansion of the Internet of Things (IoT) and heterogeneous wireless networks, the Age of Information (AoI) has emerged as a critical metric for evaluating the performance of real-time and personalized systems. While AoI-based random access is essential for next-generation applications such as the low-altitude economy and indoor service robots, existing strategies, ranging from rule-based protocols to learning-based methods, face critical challenges, including idealized model assumptions, slow convergence, and poor generalization. In this article, we propose Reflex-Core, a novel Large Language Model (LLM) agent-based framework for AoI-driven random access in heterogeneous networks. By devising an "Observe-Reflect-Decide-Execute" closed-loop mechanism, this framework integrates Supervised Fine-Tuning (SFT) and Proximal Policy Optimization (PPO) to enable optimal, autonomous access control. Based on the Reflex-Core framework, we develop a Reflexive Multiple Access (RMA) protocol and a priority-based RMA variant for intelligent access control under different heterogeneous network settings. Experimental results demonstrate that in the investigated scenarios, the RMA protocol achieves up to a 14.9% reduction in average AoI compared with existing baselines, while the priority-based version improves the convergence rate by approximately 20%.
Paper Structure (34 sections, 9 equations, 8 figures)

This paper contains 34 sections, 9 equations, 8 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. AoI Optimization
  4. LLM-based Agents in Wireless Networks
  5. System Model
  6. Network Scenarios
  7. Definition of AoI
  8. The Reflex-Core Framework
  9. The Closed-Loop Decision Mechanism
  10. Observe Module
  11. Reflect Module
  12. Decide Module
  13. Execute Module
  14. Memory Module
  15. The SFT and PPO-Based Post-Training
  16. ...and 19 more sections

Figures (8)

  • Figure 1: An illustration of the heterogeneous random access network.
  • Figure 2: Definition of the instantaneous AoI of a node over time.
  • Figure 3: An illustration of the Reflex-Core framework, which contains the main part and environment part. The main part consists of an "Observe-Reflect-Decide-Execute" closed-loop mechanism.
  • Figure 4: An illustration of the hierarchical temporal structure in Reflex-Core, showing the relationship between reflection cycles, observation periods, and time slots.
  • Figure 5: Performance comparison across different scenarios. (a)-(e) illustrate the AoI performance of RMA and LLMA methods in Scenarios 1-5, covering both homogeneous (Scenario 3) and heterogeneous (Scenarios 1-2, 4-5) network environments. (f) presents the overall performance percentage improvement, with RMA achieving 10.3-14.9% AoI reduction compared to LLMA across all scenarios.
  • ...and 3 more figures