Fresh Multiple Access: A Unified Framework Based on Large Models and Mean-Field Approximations

TL;DR

This paper introduces a unified framework for analyzing and optimizing information freshness in multi-device access scenarios using large models and mean-field approximations, addressing complexity and generality.

Contribution

It proposes a novel discrete-time tandem queue model for fresh multiple access and applies large model and mean-field techniques to handle high-dimensional analysis.

Findings

Unified framework for AoI analysis and optimization

Effective mean-field approximation reduces computational complexity

Application to age of incorrect information and peak AoII metrics

Abstract

Information freshness has attracted increasingly attention in the past decade as it plays a critical role in the emerging real-time applications. Age of information (AoI) holds the promise of effectively characterizing the information freshness, hence widely considered as a fundamental performance metric. However, in multiple-device scenarios, most existing works focus on the analysis and optimization of AoI based on queueing systems. The study for a unified approach for general multiple access control scheme in freshness-oriented scenarios remains open. In this paper, we take into consideration the combination of the fundamental freshness metric AoI and multiple access control schemes to achieve efficient cross-layer analysis and optimization in freshness-oriented scenarios, which is referred to as fresh multiple access. To this end, we build a unified framework with a discrete-time tandem queue model for fresh multiple access. The unified framework enables the analysis and optimization for general multiple access protocols in fresh multiple access. To handle the high dimension framework embedded in fresh multiple access, we introduce large model approaches for the Markov chain formulation in AoI oriented scenarios. Two typical AoI-based metric are studied including age of incorrect information (AoII) and peak AoII. Moreover, to address the computational complexity of the large model, we present mean-field approximations which significantly reduces the dimension of the Markov chain model by approximating the integral affect of massive devices in fresh multiple access.