Generative AI Enabled Matching for 6G Multiple Access

TL;DR

A framework based on generative diffusion models (GDMs) that iteratively denoises toward reward maximization to generate a matching strategy that meets specific requirements is proposed, helping to solve complex problems in 6G multiple access, such as task allocation.

Abstract

In wireless networks, applying deep learning models to solve matching problems between different entities has become a mainstream and effective approach. However, the complex network topology in 6G multiple access presents significant challenges for the real-time performance and stability of matching generation. Generative artificial intelligence (GenAI) has demonstrated strong capabilities in graph feature extraction, exploration, and generation, offering potential for graph-structured matching generation. In this paper, we propose a GenAI-enabled matching generation framework to support 6G multiple access. Specifically, we first summarize the classical matching theory, discuss common GenAI models and applications from the perspective of matching generation. Then, we propose a framework based on generative diffusion models (GDMs) that iteratively denoises toward reward maximization to generate a matching strategy that meets specific requirements. Experimental results show that, compared to decision-based AI approaches, our framework can generate more effective matching strategies based on given conditions and predefined rewards, helping to solve complex problems in 6G multiple access, such as task allocation.

Figures (5)

• Figure 1: The summary of matching generation using GenAI models: Generative Adversarial Networks, Variational Auto Encoder, Transformer and Generative Diffusion Model. Conduct an in-depth investigation of four different GenAI models from the perspectives of principles, advantages, and disadvantages.
• Figure 2: The application of matching in potential drug targets prediction, metaverse service, vehicular networks, and hybrid NOMA networks. In potential drug targets prediction, a VAE model is applied to identify the optimal matching between drugs, viruses, and hosts, effectively exploring the potential for targeted drug therapies ray2020predicting. In metaverse service, efficient matching between users and edge servers can enhance the immersive experience of users du2024diffusion. In vehicular networks, service vehicles, and base station are considered vertices, while V2V and V2I links for DNN task offloading are considered edges liu2024dnn. A structured matching strategy greatly enhances task completion efficiency. In multiple access wireless networks, decision-making AI algorithms such as DRL interact with the dynamic environment and maximize the reward function to achieve effective matching between physical entities and wireless resources, thereby improving system spectral efficiency and throughput 93529569964376.
• Figure 3: The framework of the proposed matching method. In part A, the framework follows four steps. Step 1, the random noise is mapped to a noisy graph through one-hot encoding. Step 2, the condition, denoising step, and the noisy graph containing matching relationships are input into the denoising network to iteratively generate the matching strategies. Step 3, multiple generated trajectories are sampled from the denoising network, and rewards are calculated to obtain the sum of gradients for model parameter updating. Finally, the denoising network which is iteratively updated can generate the optimal matching strategy based on the current environment. Part B illustrates the RSMA-aided AIGC service provider selection problem in Section \ref{['case_study']}.
• Figure 4: The training curve of the proposed GDM-based matching method and the comparison with other methods.
• Figure 5: Normalized QoE versus SNR.