Mobile-Edge Computing (MEC) could relieve computing pressure and save energy of resource-constrained Smart Mobile Devices (SMDs) via computation offloading. Nevertheless, offloading strategy design for multiuser MEC systems is challenging. Specifically, offloading operations (i.e., terminal execution strategy, access rate, and cloud execution strategy) are not only inner-coupled for each SMD due to parallel local and cloud execution, but also inter-coupled among SMDs due to competition for radio and computation resources. Worse still, the inner- and inter-coupling interplay each other. However, existing works generally weaken this inner-inter-coupling, resulting in an inability to adapt to network differences, terminal capacity differences, and application requirements differences. Hence, only suboptimal performance could be achieved. As motivated, we jointly optimizes terminal execution strategy, radio resource allocation, and MEC computation resource allocation to minimize weighted sum of terminal energy consumption. Additionally, via dynamically matching individual offloading behavior and group’s competitive resources allocation, our proposed algorithm could not only reflect mechanism of interaction between inner- and inter-coupling relationship, but also well adopt to diversities of network conditions, terminal capacity, and application requirements to further harvest MEC gain. Finally, simulation results demonstrate that our algorithm significantly outperforms existing schemes, more specifically up to 73.8% less energy consumption.

中文翻译：

---

与终端、无线电接入网络和边缘服务器协作的节能多用户部分计算卸载

移动边缘计算 (MEC) 可以通过计算卸载减轻资源受限的智能移动设备 (SMD) 的计算压力并节省能源。然而，多用户 MEC 系统的卸载策略设计具有挑战性。具体来说，卸载操作（即终端执行策略、访问速率和云执行策略）不仅由于本地和云端并行执行而对每个 SMD 内部耦合，而且由于无线电和计算的竞争，SMD 之间也相互耦合资源。更糟糕的是，内部耦合和内部耦合相互影响。然而，现有的工作普遍削弱了这种内部相互耦合，导致无法适应网络差异、终端​​容量差异和应用需求差异。因此，只能实现次优的性能。作为激励，我们联合优化终端执行策略、无线资源分配和MEC计算资源分配，以最小化终端能耗的加权和。此外，通过动态匹配个体卸载行为和群体竞争资源分配，我们提出的算法不仅可以反映内耦合和互耦合关系之间的交互机制，而且可以很好地适应网络条件、终端容量和应用需求的多样性。进一步收获MEC收益。最后，仿真结果表明，我们的算法明显优于现有方案，更具体地说，能耗降低了 73.8%。和 MEC 计算资源分配，以最小化终端能耗的加权和。此外，通过动态匹配个体卸载行为和群体竞争资源分配，我们提出的算法不仅可以反映内耦合和互耦合关系之间的交互机制，而且可以很好地适应网络条件、终端容量和应用需求的多样性。进一步收获MEC收益。最后，仿真结果表明，我们的算法明显优于现有方案，更具体地说，能耗降低了 73.8%。和 MEC 计算资源分配，以最小化终端能耗的加权和。此外，通过动态匹配个体卸载行为和群体竞争资源分配，我们提出的算法不仅可以反映内耦合和互耦合关系之间的交互机制，而且可以很好地适应网络条件、终端容量和应用需求的多样性。进一步收获MEC收益。最后，仿真结果表明，我们的算法明显优于现有方案，更具体地说，能耗降低了 73.8%。我们提出的算法不仅可以反映内耦合和互耦合关系之间的交互机制，而且可以很好地适应网络条件、终端容量和应用需求的多样性，以进一步收获 MEC 增益。最后，仿真结果表明，我们的算法明显优于现有方案，更具体地说，能耗降低了 73.8%。我们提出的算法不仅可以反映内耦合和互耦合关系之间的交互机制，而且可以很好地适应网络条件、终端容量和应用需求的多样性，以进一步收获 MEC 增益。最后，仿真结果表明，我们的算法明显优于现有方案，更具体地说，能耗降低了 73.8%。