An online dual consensus algorithm for distributed resource allocation over networks

被引：0

作者：

Chen, Yuwei ^{[1
]}

Deng, Zengde ^{[1
]}

Yuan, Biao ^{[4
]}

Chen, Zaiyi ^{[1
]}

Chen, Yujie ^{[2
]}

Hu, Haoyuan ^{[3
]}

机构：

[1] Cainiao Network, Hangzhou, Zhejiang, Peoples R China

[2] Cainiao Network, Applicat Algorithm Design & Dev, Hangzhou, Zhejiang, Peoples R China

[3] Cainiao Network, Artificial Intelligence Dept, Hangzhou, Zhejiang, Peoples R China

[4] Shanghai Jiao Tong Univ, Sino US Global Logist Inst, Data Driven Management Decis Making Lab, Shanghai, Peoples R China

来源：

IISE TRANSACTIONS | 2024年

基金：

中国国家自然科学基金;

关键词：

Distributed decision-making; online optimization; resource allocation; ADMM; OPTIMIZATION;

D O I：

10.1080/24725854.2024.2428652

中图分类号：

T [工业技术];

学科分类号：

08 ;

摘要：

We address the problem of online resource allocation in a distributed environment where requests arrive dynamically over time at different agents in the network. As each request arrives, the receiving agent must make an immediate decision that incurs a cost and consumes a certain amount of resources. The requests are drawn independently from unknown distributions that are different for each agent. First, we present an Online Consensus Alternating Direction Method of Multipliers (OC-ADMM) algorithm for the dual counterpart of the online distributed resource allocation problem, focusing on the dual variables. Then, we propose an Online Dual Consensus ADMM (ODC-ADMM) algorithm for the primal problem to derive the primal variables from the dual update process in the OC-ADMM algorithm. The ODC-ADMM algorithm exhibits sublinear growth in both regret and expected constraint violation with respect to the time horizon. Furthermore, extensive numerical results on both synthetic and real-world data confirm its effectiveness.

引用

页数：12