.png){kind=icon}
.png){kind=icon}
Return
Cross-layer dual-attention-based priority task scheduling for cloud-edge-end computing
.png){kind=icon}
PRE
AI
DOI：10.1007/s11227-026-08654-8.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Cloud-edge-end computing provides favorable support for emerging applications by virtue of its low-latency and high reliability. However, efficient priority-aware task scheduling remains challenging because task loads, resource availability, and deadline requirements vary dynamically. This work focuses on independent priority-aware computational task scheduling in a controlled cloud-edge-end inspired environment, where tasks have heterogeneous resource demands, execution times, and deadlines. To address this problem, a Cross-layer Dual-Attention-based Hierarchical Deep Reinforcement Learning method (CDAHDRL) is proposed. The upper layer uses a proximal policy optimization (PPO) agent to adaptively select a priority dispatching rule for global task ordering, while the lower layer uses an Actor-Critic agent to assign the selected task to a computing node. A cross-layer dual-attention module is introduced to exchange information between global priority selection and local resource allocation. Experimental results under controlled simulation settings show that CDAHDRL achieves more stable overall performance than representative heuristic, flat reinforcement learning, staged scheduling, and no-attention variants in terms of makespan, average delay, and on-time completion rate.
Keywords:
Cloud-edge-end computing
Task scheduling
Cross-layer dual-attention mechanism
Deep reinforcement learning
Journal
T
IF:
0
Papers:
647
Citations:
0