终端直通辅助的移动边缘计算任务动态卸载方案

D2D-assisted Dynamic Task Offloading Scheme in Mobile Edge Computing

针对密集型应用任务面临着网络服务资源受限且无法满足日益增长的用户需求这一问题,提出了一种终端直通(Device-to-Device,D2D)辅助的移动边缘计算(Mobile Edge Computing,MEC)任务卸载架构,利用D2D协作中继技术,可将计算量大的任务通过D2D链路卸载到远端MEC服务器进行处理。其研究目标是为每个任务选择最优的卸载策略实现系统能耗最小化,即在本地执行任务,直接卸载到D2D设备,或在最合适的D2D设备帮助下卸载到MEC服务器。考虑到边缘网络的高动态特性,设计了D2D协同的边缘网络系统模型,并在此基础上构建了受时延约束的端到端优化目标函数。利用深度强化学习自学习的优势,将任务卸载决策问题建模为马尔可夫模型,并采用双深度Q网络(Double DQN,DDQN)算法对问题进行求解。仿真数据表明,提出的D2D协同计算方案较其他算法能有效降低移动用户的任务执行能耗。

Intensive application tasks are faced with the problem that limited network service resources cannot meet the ever-increasing demands of mobile users.To address this issue,a task offloading framework for Device-to-Device(D2D)-enabled Mobile Edge Computing(MEC)is proposed.By introducing the D2D cooperative relay technology,tasks with large calculation amount are offloaded to the MEC server at the network edge through D2D links for data processing.The research goal is to minimize the system energy consumption by optimizing the offloading decisions,namely,whether to perform the task locally and directly offload it to the D2D device,or to offload the task to the MEC server with the help of the most suitable D2D device.Considering the high dynamic characteristics of the edge network,a D2D collaborative edge network system model is designed and an end-to-end optimization objective function subjected to delay constraints is formulated.Secondly,the problem of task offloading decision-making is modeled as a Markov model,and the problem is efficiently solved by the Double DQN(DDQN)algorithm.Simulation results show that the proposed D2D collaborative computing scheme can significantly reduce the energy consumption of task execution of mobile users compared with other algorithms.