Adaptive Multiple Access and Service Placement for Generative Diffusion Models

2025-08-11

AI Summary

To address the high latency and resource overhead associated with deploying generative diffusion models (GDMs) in mobile edge networks, this paper proposes a joint service placement and multi-user access optimization framework. Methodologically, it (1) dynamically partitions denoising blocks across heterogeneous edge nodes; (2) introduces an adaptive inference step reduction mechanism to alleviate both communication and computational loads; and (3) designs a Double Dueling Deep Q-Network (D3QN)-based service placement algorithm, integrated with a greedy access policy and the LEARN-GDM framework—a deep reinforcement learning-driven orchestration scheme. Experimental results demonstrate that, compared to monolithic and fixed-chain deployments, the proposed approach significantly improves scalability and latency robustness while maintaining quality-of-service guarantees, thereby substantially enhancing the efficiency of edge-hosted GDM inference.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Diffusion Models for Vision

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Generative Diffusion Models (GDMs) have emerged as key components of Generative Artificial Intelligence (GenAI), offering unparalleled expressiveness and controllability for complex data generation tasks. However, their deployment in real-time and mobile environments remains challenging due to the iterative and resource-intensive nature of the inference process. Addressing these challenges, this paper introduces a unified optimization framework that jointly tackles service placement and multiple access control for GDMs in mobile edge networks. We propose LEARN-GDM, a Deep Reinforcement Learning-based algorithm that dynamically partitions denoising blocks across heterogeneous edge nodes, while accounting for latent transmission costs and enabling adaptive reduction of inference steps. Our approach integrates a greedy multiple access scheme with a Double and Dueling Deep Q-Learning (D3QL)-based service placement, allowing for scalable, adaptable, and resource-efficient operation under stringent quality of service requirements. Simulations demonstrate the superior performance of the proposed framework in terms of scalability and latency resilience compared to conventional monolithic and fixed chain-length placement strategies. This work advances the state of the art in edge-enabled GenAI by offering an adaptable solution for GDM services orchestration, paving the way for future extensions toward semantic networking and co-inference across distributed environments.

Problem

• Optimize service placement for Generative Diffusion Models in mobile edge networks
• Develop adaptive multiple access control for resource-efficient GDM deployment
• Reduce latency and improve scalability in real-time GDM inference

Innovation

• Deep Reinforcement Learning-based dynamic denoising block partitioning
• Greedy multiple access with D3QL-based service placement
• Adaptive reduction of GDM inference steps