PhD Forum Abstract: Diffusion-based Task Scheduling for Efficient AI-Generated Content in Edge Networks
Changfu Xu · 2024
The Artificial Intelligence-Generated Content (AIGC) technique has gained significant popularity in creating diverse content. However, the current deployment of AIGC services is a centralized framework, thus leading to high response times. To address this issue, we propose a diffusion-based task scheduling method that considers the integration of the diffusion model, Deep Reinforcement Learning (DRL), and Mobile Edge Computing (MEC) technique to improve the AIGC efficiency. This challenges efficient server selection without prior information in dynamic MEC systems. We formulate our problem as an online integer linear programming problem aiming to minimize task offloading delay. Furthermore, we propose a novel AIGC Task Scheduling (DDRL-ATS) algorithm based on Diffusion DRL (DDRL) that effectively addresses this problem. The DDRL-ATS algorithm achieves efficient AIGC tailored for heterogeneous MEC environments. Additionally, an online Adaptive Multi-server Selection and Allocation (DDRL-AMSA) algorithm based on DDRL is proposed to further enhance the AIGC efficiency. Moreover, our DDRL-AMSA algorithm achieves near-optimal solutions within approximate linear time complexity bounds. Finally, experimental results validate the effectiveness of our method by showcasing at least a reduction of 13.54% in task offloading delay compared to state-of-the-art methods.