【摘要】 MEC网络将计算和存储资源下沉到网络边缘，以满足未来6G业务的低时延要求，MEC网络中多维资源的管理和调度决策是提升用户体验的关键。针对MEC网络中的资源管理决策问题，提出了基于Lyapunov优化的分布式资源管理算法。具体来讲，引入任务数据队列及虚拟能量队列，以确保任务执行的公平性，且避免了系统负载过大时造成的过度拥塞问题，并且通过Lyapunov优化理论构建目标函数，基于DDPG算法进行求解。另外，考虑MEC服务器对基于DDPG的决策模型具有不同需求，将异构网络部署在不同的边缘服务器中，构建分布式多连续变量决策模型。仿真结果表明，所使用的分布式决策算法的收敛性和稳定性更优。更多还原

【Abstract】 Mobile edge computing(MEC) networks bring computation and storage resources closer to the network edge to meet the low-latency requirements of future 6G services. Efficient management and scheduling of multidimensional resources in MEC networks are key to enhancing user experience. To address the resource management decision-making problem in MEC networks, this study proposes a distributed resource management algorithm based on Lyapunov optimization. Specifically, task data queues and virtual energy queues are introduced to ensure task execution fairness and mitigate excessive congestion under high system loads. The objective function is constructed using Lyapunov optimization theory and solved with a deep deterministic policy gradient(DDPG)-based algorithm. Furthermore, considering that MEC servers have varying requirements for DDPG-based decision models, heterogeneous networks are deployed across different edge servers, creating a distributed multi-continuous variable decision model. Simulation results demonstrate the superior convergence and stability of the proposed distributed decision-making algorithm.更多还原