Edge computing (EC) provides an effective means to cope with explosive computation demands of the Internet-of-Things (IoT). This paper presents a new cooling-aware joint optimization of the CPU configuration of the edge servers, and the schedules of wireless power transfer (WPT), offloading and computing for WPT-powered devices, so that the resource-restrained devices can have tasks accomplished in a timely and energy-efficient manner. Alternating optimization is applied to minimize the total energy consumption of WPT, EC, and cooling, while satisfying the computation deadlines of the devices. A key aspect is that semi-closed-form solutions are derived for the WPT power, offloading duration, and CPU frequency by applying the Lagrange duality method. With the solutions, the alternating optimization converges quickly and indistinguishably closely to the lower bound of the energy consumption. The semi-closed-form solutions also reveal the structure underlying the optimal solution to the problem, and can validate the result of the alternating optimization. Extensive simulations show that the proposed algorithm can save up to 90.4% the energy of existing benchmarks in our considered cases.

中文翻译：

---

无线供电设备边缘服务器配置和边缘计算卸载的冷却感知优化

边缘计算（EC）提供了一种有效的手段来应对物联网（IoT）的爆炸性计算需求。本文提出了一种新的冷却感知联合优化边缘服务器的 CPU 配置，以及无线功率传输 (WPT)、WPT 供电设备的卸载和计算时间表，以便资源受限的设备可以完成任务及时和节能的方式。应用交替优化以最小化 WPT、EC 和冷却的总能耗，同时满足设备的计算期限。一个关键方面是通过应用拉格朗日对偶方法为 WPT 功率、卸载持续时间和 CPU 频率推导出半封闭形式的解决方案。有了解决方案，交替优化快速且难以区分地收敛到能源消耗的下限。半封闭形式的解还揭示了问题最优解的结构，可以验证交替优化的结果。大量模拟表明，在我们考虑的情况下，所提出的算法可以节省高达 90.4% 的现有基准的能量。