基于MADDPG的边缘网络任务卸载与资源管理

MADDPG-based Task Offloading and Resource Management for Edge Networks

随着物联网的飞速发展,连接到第六代无线移动网络(6G)的智能设备数量急剧增加。由于存在多维网络资源并存,网络设备异构,网络结构复杂时变等一系列问题,无线网络面临着前所未有的挑战。基于边缘设备的新型网络提供了低时延的就近处理计算和通信资源分配等问题,是合理解决巨量智能设备接入下通信资源分配的有效解决方案。将AI技术融入边缘计算网络架构中,提出了一种基于深度确定性策略梯度算法(MADDPG)的边缘计算网络任务卸载与资源管理模型,通过联合优化任务分层卸载和资源分配,实现处理效率的最大化。

With the rapid development of IoT,the number of smart devices connected to the sixth generation wireless mobile network(6G)has increased dramatically.Due to the coexistence of multi-dimensional network resources,heterogeneous network equipment,and complex and time-varying network structures,wireless networks are facing unprecedented challenges.The new network based on edge devices provides low latency proximity to deal with problems such as computation and communication resource allocation,and is an effective solution to reasonably solve the communication resource allocation under the access of huge number of smart devices.In this paper,AI technology is incorporated into the edge computing network architecture,and a task offloading and resource management model based on deep deterministic policy gradient algorithm(MADDPG)for edge computing networks is proposed to maximize processing efficiency by jointly optimizing task hierarchical offloading and resource allocation.