.png){kind=icon}
.png){kind=icon}
Return
DAMRL: A dependency-aware meta reinforcement learning framework for adaptive task offloading in vehicular edge computing
.png){kind=icon}
PRE
AI
Z
Y
B
X
DOI：10.1016/j.future.2026.108740.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Proposes a dependency-aware and fast-adaptive task offloading framework for DAG tasks in highly dynamic vehicular edge computing. • Designs a multi-head graph attention network to generate task embeddings for dependency-aware offloading decisions, thereby enhancing parallelism and reducing task execution time. • Employs a context-based meta reinforcement learning enhanced with a dual replay buffer mechanism, enabling fast, sample-efficient, and stable policy adaptation to new scenarios. • Achieves faster adaptation to new scenarios and superior performance in terms of computation waiting time, load balancing ratio and fairness of vehicles compared with four representative baselines.
Journal
F
IF:
6.1
Papers:
6.8K
Citations:
2.3W
Organization
No organization information available
Cited Papers