Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Semantic-Aware Multiple Access Scheme for Distributed, Dynamic 6G-Based Applications

Hamidreza Mazandarani, Masoud Shokrnezhad, Tarik Taleb

TL;DR

This work introduces a semantic-aware formulation for distributed 6G MAC and proposes SAMA-D3QL, a model-free multi-agent deep reinforcement learning framework that enables UEs to autonomously access wireless spectrum while accounting for data correlations through assisted throughput. By decomposing UE throughput into self- and assisted components and optimizing an $\alpha$-fairness objective, the approach addresses utilization–fairness trade-offs in dynamic, distributed environments. The method employs a centralized training, decentralized execution paradigm with a VDN-based structure and a D3QL target, demonstrated to outperform semantic-oblivious baselines and show robustness across single- and multi-channel scenarios. These results suggest that semantic-aware MADRL can significantly enhance resource allocation for federated, evolving 6G applications, with potential extensions to more general association models and realistic channel dynamics.

Abstract

The emergence of the semantic-aware paradigm presents opportunities for innovative services, especially in the context of 6G-based applications. Although significant progress has been made in semantic extraction techniques, the incorporation of semantic information into resource allocation decision-making is still in its early stages, lacking consideration of the requirements and characteristics of future systems. In response, this paper introduces a novel formulation for the problem of multiple access to the wireless spectrum. It aims to optimize the utilization-fairness trade-off, using the $α$-fairness metric, while accounting for user data correlation by introducing the concepts of self- and assisted throughputs. Initially, the problem is analyzed to identify its optimal solution. Subsequently, a Semantic-Aware Multi-Agent Double and Dueling Deep Q-Learning (SAMA-D3QL) technique is proposed. This method is grounded in Model-free Multi-Agent Deep Reinforcement Learning (MADRL), enabling the user equipment to autonomously make decisions regarding wireless spectrum access based solely on their local individual observations. The efficiency of the proposed technique is evaluated through two scenarios: single-channel and multi-channel. The findings illustrate that, across a spectrum of $α$ values, association matrices, and channels, SAMA-D3QL consistently outperforms alternative approaches. This establishes it as a promising candidate for facilitating the realization of future federated, dynamically evolving applications.

A Semantic-Aware Multiple Access Scheme for Distributed, Dynamic 6G-Based Applications

TL;DR

This work introduces a semantic-aware formulation for distributed 6G MAC and proposes SAMA-D3QL, a model-free multi-agent deep reinforcement learning framework that enables UEs to autonomously access wireless spectrum while accounting for data correlations through assisted throughput. By decomposing UE throughput into self- and assisted components and optimizing an -fairness objective, the approach addresses utilization–fairness trade-offs in dynamic, distributed environments. The method employs a centralized training, decentralized execution paradigm with a VDN-based structure and a D3QL target, demonstrated to outperform semantic-oblivious baselines and show robustness across single- and multi-channel scenarios. These results suggest that semantic-aware MADRL can significantly enhance resource allocation for federated, evolving 6G applications, with potential extensions to more general association models and realistic channel dynamics.

Abstract

The emergence of the semantic-aware paradigm presents opportunities for innovative services, especially in the context of 6G-based applications. Although significant progress has been made in semantic extraction techniques, the incorporation of semantic information into resource allocation decision-making is still in its early stages, lacking consideration of the requirements and characteristics of future systems. In response, this paper introduces a novel formulation for the problem of multiple access to the wireless spectrum. It aims to optimize the utilization-fairness trade-off, using the -fairness metric, while accounting for user data correlation by introducing the concepts of self- and assisted throughputs. Initially, the problem is analyzed to identify its optimal solution. Subsequently, a Semantic-Aware Multi-Agent Double and Dueling Deep Q-Learning (SAMA-D3QL) technique is proposed. This method is grounded in Model-free Multi-Agent Deep Reinforcement Learning (MADRL), enabling the user equipment to autonomously make decisions regarding wireless spectrum access based solely on their local individual observations. The efficiency of the proposed technique is evaluated through two scenarios: single-channel and multi-channel. The findings illustrate that, across a spectrum of values, association matrices, and channels, SAMA-D3QL consistently outperforms alternative approaches. This establishes it as a promising candidate for facilitating the realization of future federated, dynamically evolving applications.
Paper Structure (14 sections, 10 equations, 4 figures, 2 tables, 1 algorithm)

This paper contains 14 sections, 10 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Proposed Method
  4. Optimal Solutions
  5. Learning Mechanism
  6. Action Space
  7. State Space
  8. Reward
  9. Training Process
  10. Evaluation
  11. Scenario 1: Single-Channel Setting
  12. Scenario 2: Multi-channel Setting
  13. Conclusion
  14. General Setting of the Problem

Figures (4)

  • Figure 1: An illustration of similar segments.
  • Figure 2: A sample network consisting of four UEs ($\mathcal{N} = 4$) and five segments ($\mathcal{K} = 5$). As illustrated, one separate segment is associated with each user, in addition to one shared segment among the first and second UEs.
  • Figure 3: Objective functions over time and all-time average UE throughputs in the first scenario, encompassing experiments A and B, comparing the SAMA-D3QL, MA-D3QL, and RND (random agents) algorithms. The reported results are based on averaging the outcomes of 10 rounds of simulations. It's noteworthy that, when present, the lighter segment of each UE throughput signifies the assisted throughput for that particular UE.
  • Figure 4: The average value of objective functions for the last 1000 time slots (of 10k time slots) in experiments A, B, and C. The UE-segment association matrix for each experiment is illustrated. Results are the average of 10 rounds of simulations.