.png){kind=icon}
Return
Joint Task Coding and Transfer Optimization for Edge Computing Power Networks
.png){kind=icon}
PRE
AI
DOI：10.1109/TNSE.2025.3554100.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Driven by the exponential growth of the Internet of Everything (IoE) and substantial advancements in artificial intelligence, services based on deep learning have seen a significant increase in demand for computing resources. The existing edge computing paradigms struggle to handle the explosive growth in computing demands. They also face challenges in jointly optimizing the high transmission load and privacy concerns of task collaboration while failing to utilize computing resources efficiently in complex and dynamic computing power networks. In this paper, we investigate an edge computing power network framework that integrates heterogeneous computing resources from both horizontal and vertical dimensions. We formulate a collaborative task transfer problem to minimize the total execution time of multiple tasks by joint optimization task coding, computing-task association, and collaborative transfer computing strategies among nodes. To solve the formulated problem, we conduct in-depth theoretical analyses and design a two-layer multi-agent optimization algorithm. Specifically, the task coding problem is reformulated in the inner layer into a solvable form, and a closed-form expression for the task coding ratio is derived. Subsequently, we design an adaptive hybrid reward-based multi-agent deep reinforcement learning algorithm to address the sparsity challenges of single-layer rewards while ensuring efficient and stable training convergence. Numerical results show the superiority of our proposed algorithm.
Keywords:
Edge intelligence
computing power networks
task coding
multi-agent deep reinforcement learning
Journal
I
IF:
7.9
Papers:
2.6K
Citations:
10.0K
Organization
View more organizations
Cited Papers
.png){kind=icon}
ShareSave
ShareSaveShareSave
ShareSaveShareSave View more