arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1902.08779cs.ITmath.IT
收藏

具有动态任务到达的无线移动边缘计算最优资源分配

Optimal Resource Allocation for Wireless Powered Mobile Edge Computing with Dynamic Task Arrivals

Feng Wang, Hong Xing, Jie Xu

摘要:

本文考虑了一种无线供电的多用户移动边缘计算(MEC)系统,其中多天线接入点(AP)采用基于射频(RF)信号的无线功率传输(WPT)对多个分布式用户进行充电,每个用户利用收获的能量通过本地计算和任务卸载来执行计算任务。我们认为频分多址(FDMA)协议,以支持同时从多个用户到AP的任务卸载。不同于以往的作品,认为一次性优化与静态任务模型,我们研究了联合计算和无线资源分配优化与动态任务到达在一个有限的时间范围内由多个时隙。在这种设置下,我们的目标是最大限度地减少系统的能量消耗,包括AP的传输能量和MEC服务器的计算能量在整个地平线上,通过联合优化在AP的发射能量波束成形,并在不同的时隙在用户的本地计算和任务卸载策略。为了表征这样的系统的基本性能限制,我们专注于离线优化,假设任务和信道信息是已知的先验在AP。在这种情况下,能量最小化问题对应于凸优化问题。利用拉格朗日对偶方法,我们得到了井结构中该问题的最优解。结果表明,为了最大限度地提高系统的能量效率,在每个用户和AP的任务输入比特的最佳数量随时间单调增加,在不同的用户的卸载策略取决于无线信道条件和任务负载在AP。数值结果表明,拟议的联合WPT-MEC设计的好处,替代基准计划没有这样的联合设计。

英文摘要:

This paper considers a wireless powered multiuser mobile edge computing (MEC) system, where a multi-antenna access point (AP) employs the radio-frequency (RF) signal based wireless power transfer (WPT) to charge a number of distributed users, and each user utilizes the harvested energy to execute computation tasks via local computing and task offloading. We consider the frequency division multiple access (FDMA) protocol to support simultaneous task offloading from multiple users to the AP. Different from previous works that considered one-shot optimization with static task models, we study the joint computation and wireless resource allocation optimization with dynamic task arrivals over a finite time horizon consisting of multiple slots. Under this setup, our objective is to minimize the system energy consumption including the AP's transmission energy and the MEC server's computing energy over the whole horizon, by jointly optimizing the transmit energy beamforming at the AP, and the local computing and task offloading strategies at the users over different time slots. To characterize the fundamental performance limit of such systems, we focus on the offline optimization by assuming the task and channel information are known a-priori at the AP. In this case, the energy minimization problem corresponds to a convex optimization problem. Leveraging the Lagrange duality method, we obtain the optimal solution to this problem in a well structure. It is shown that in order to maximize the system energy efficiency, the optimal number of task input-bits at each user and the AP are monotonically increasing over time, and the offloading strategies at different users depend on both the wireless channel conditions and the task load at the AP. Numerical results demonstrate the benefit of the proposed joint-WPT-MEC design over alternative benchmark schemes without such joint design.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑