Multi-access mobile edge computing (MEC) enables mobile users (MUs) to offload tasks to proximal multiple MEC servers for fast task processing. Since MUs generally have stringent delay requirements and limited energy and MEC servers have finite communication, computation, and caching (3C) resources, the joint co-design and optimization over computation offloading and 3C resource allocation for wireless-powered multi-access MEC systems have been highly demanded, while having not been well studied yet. We consider a wireless-powered multi-access MEC, where multiple energy harvesting (EH) based MUs and multiple base stations (BSs) each equipped with an MEC server coexist. Each MU first harvests energy from nearby MEC servers, and then offloads its sub-tasks to multiple MEC servers concurrently based on non-orthogonal multiple access (NOMA), so as to reduce data offloading delay. We first formulate a delay minimization problem, by jointly optimizing MUs’ computation offloading, MEC servers’ computation and caching resources allocation, and system communication resources allocation. Then we propose an alternating direction multiplier method (ADMM) based distributed scheme to decompose the formulated optimization problem into several sub-problems, and use the block coordinate descending (BCD) method and the successive convex approximation (SCA) method to transform all sub-problems each corresponding to one MEC server to convex subproblems. Finally, we validate and evaluate our proposed scheme through numerical analyses, which show that our proposed distributed scheme can greatly reduce the min–max delay of all MUs by allowing each MU’s concurrent data offloading to multiple MEC servers using NOMA and by jointly optimizing computation offloading and 3C resource allocations. Also, the delay imposed by our proposed scheme is much smaller than that imposed by the Interior Point method, which shows the effectiveness of our proposed scheme.

中文翻译：

---

通过无线供电和支持 NOMA 的多路访问 MEC 进行计算卸载和 3C 资源分配的联合优化

多接入移动边缘计算 (MEC) 使移动用户 (MU) 能够将任务卸载到邻近的多个 MEC 服务器，以实现快速任务处理。由于MU通常具有严格的延迟要求和有限的能量，并且MEC服务器具有有限的通信、计算和缓存（3C）资源，因此无线供电的多路访问MEC系统的计算卸载和3C资源分配的联合协同设计和优化已经成为可能。需求量很大，但尚未得到充分研究。我们考虑一种无线供电的多接入MEC，其中多个基于能量收集（EH）的MU和多个基站（BS）共存，每个基站都配备了MEC服务器。每个MU首先从附近的MEC服务器获取能量，然后基于非正交多路访问（NOMA）将其子任务同时卸载到多个MEC服务器，以减少数据卸载延迟。我们首先通过联合优化 MU 的计算卸载、MEC 服务器的计算和缓存资源分配以及系统通信资源分配，提出延迟最小化问题。然后，我们提出一种基于交替方向乘子法（ADMM）的分布式方案，将公式化的优化问题分解为多个子问题，并使用块坐标下降（BCD）方法和逐次凸逼近（SCA）方法来变换所有子问题每个问题对应一个MEC服务器来凸子问题。最后，我们通过数值分析验证和评估我们提出的方案，这表明我们提出的分布式方案可以通过允许每个 MU 使用 NOMA 并发数据卸载到多个 MEC 服务器并联合优化计算卸载，从而大大减少所有 MU 的最小-最大延迟和3C资源分配。此外，我们提出的方案所施加的延迟远小于内点方法所施加的延迟，这表明了我们提出的方案的有效性。