计算机工程2025，Vol.51Issue(5)：83-92,10.DOI:10.19678/j.issn.1000-3428.0068958
基于多智能体深度强化学习的SD-IoT控制器部署
SD-IoT Controller Placement Based on Multi-Agent Deep Reinforcement Learning
摘要
Abstract
The rapid growth of Internet of Things(IoT)traffic has significantly impacted data transmission for devices such as sensors.Software-Defined Networking(SDN)offers a solution to optimize network performance and enhance data transmission quality.However,the dynamic nature of network states,such as traffic fluctuations in IoT environments,poses challenges to the performance of the control plane in SDN.This study addresses the dynamic controller placement problem in Software-Defined IoT(SD-IoT)to ensure consistent control plane performance under changing traffic conditions.The approach considers the energy consumption and data transmission characteristics of IoT nodes when deploying controllers,with a comprehensive evaluation of factors such as delay,control reliability,and energy consumption.The problem is modeled as a Markov game process to capture these dynamics effectively.To optimize both individual controller performance and the overall control plane performance,this study employs multi-agent deep reinforcement learning.During the deployment phase,action masks are utilized to exclude nodes with insufficient performance or limited power supply,ensuring robust and efficient controller placement.Simulation experiments demonstrate that the proposed algorithm identifies high-performance deployment solutions compared with the placement algorithms based on Louvain community division or single agent Deep Q-Network(DQN),achieving superior results in dynamic IoT environments.关键词
软件定义物联网/控制器部署/多智能体深度强化学习/动作掩码/马尔可夫博弈Key words
Software-Defined Internet of Things(SD-IoT)/controller placement/multi-agent deep reinforcement learning/action mask/Markov game分类
计算机与自动化引用本文复制引用
吕超峰,徐鹏飞,罗迪,刘金平..基于多智能体深度强化学习的SD-IoT控制器部署[J].计算机工程,2025,51(5):83-92,10.基金项目
湖南省教育厅科学研究项目(23C1042). （23C1042）
