Article: Multi-agent deep reinforcement learning edge task scheduling algorithm with migratable service environment Journal: International Journal of Sensor Networks (IJSNET) 2026 Vol.50 No.1 pp.1 - 18 Abstract: The multi-edge collaborative computing approach stores task service environments in edge nodes closer to end-users and uses multi-edge networks for collaborative offloading, overcoming long transmission distances and slow response times in traditional cloud computing. However, existing fixed-storage task offloading methods cannot dynamically schedule service environments for regional task preferences, leading to unbalanced multi-edge loads and reduced execution efficiency. We propose a container-based migratable service environment scheduling model to dynamically meet regional service demands via real-time task and environment scheduling. To address process coupling and storage replacement issues, we integrate offloading, environment migration, and content replacement into a unified scheduling action using reinforcement learning. Our improved multi-agent deep RL algorithm employs centralised training with distributed execution and an attention mechanism to optimise policy learning. Simulations show the approach enhances multi-edge load balancing and reduces average task delay. Inderscience Publishers - linking academia, business and industry through research

Title: Multi-agent deep reinforcement learning edge task scheduling algorithm with migratable service environment

Authors: Zengwei Lyu; Yu Zhang; Zhenchun Wei; Juan Xu; Lei Shi; Yuqi Fan

Addresses: School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, 230009, China ' School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, 230009, China ' School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, 230009, China ' School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, 230009, China ' School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, 230009, China ' School of Computer Science and Information Engineering, Hefei University of Technology, Hefei, 230009, China

Abstract: The multi-edge collaborative computing approach stores task service environments in edge nodes closer to end-users and uses multi-edge networks for collaborative offloading, overcoming long transmission distances and slow response times in traditional cloud computing. However, existing fixed-storage task offloading methods cannot dynamically schedule service environments for regional task preferences, leading to unbalanced multi-edge loads and reduced execution efficiency. We propose a container-based migratable service environment scheduling model to dynamically meet regional service demands via real-time task and environment scheduling. To address process coupling and storage replacement issues, we integrate offloading, environment migration, and content replacement into a unified scheduling action using reinforcement learning. Our improved multi-agent deep RL algorithm employs centralised training with distributed execution and an attention mechanism to optimise policy learning. Simulations show the approach enhances multi-edge load balancing and reduces average task delay.

Keywords: multi-agent reinforcement learning; edge computing; task scheduling; edge computing; migratable service environment.

DOI: 10.1504/IJSNET.2026.151234

International Journal of Sensor Networks, 2026 Vol.50 No.1, pp.1 - 18

Received: 17 Mar 2024
Accepted: 08 Jul 2025
Published online: 19 Jan 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Multi-agent deep reinforcement learning edge task scheduling algorithm with migratable service environment

Keep up-to-date