Multi-access mobile edge computing (MEC) allows each user to offload partial task bits to multiple MEC servers via multi-access radio transmission, enabling the collaboration of edge computing. By leveraging the advantage of nonorthogonal multiple access (NOMA) in improving the transmission efficiency, it is expected to effectively reduce the energy consumption of users in multi-access MEC with the aid of NOMA. However, considering the co-channel interference of NOMA and computation collaboration among MEC servers, the joint task offloading and resource allocation is a challenging problem. In this article, with the objective to minimize the weighted sum energy of users, we first propose optimal task offloading and resource allocation algorithms for the special single-user (OTORA-SU) and general multi-user (OTORA-MU) cases, by exploiting the convex and layered structure of the formulated problem. Interestingly, it is found that the single user always preferentially offloads task bits to the MEC servers with better channel gains, regardless of the computation capacity of each MEC server. Then, a low-complexity algorithm (LTORA-MU) is proposed for the general multi-user case, which converges fast and achieves near-optimal performances. Considering the channel estimation error, the impacts of imperfect channel state information (CSI) and successive interference cancellation (SIC) are also investigated. Simulation results demonstrate that for both perfect and imperfect CSI and SIC, 1) the proposed OTORA-SU and LTORA-MU outperform the local computing, full offloading, FDMA-based offloading and non-collaborative MEC schemes; 2) as the number of MEC servers grows, the ratio of offloading task bits increases while the energy consumption is decreased.

中文翻译：

---

启用NOMA的多路访问移动边缘计算的联合任务分载和资源分配

多访问移动边缘计算（MEC）允许每个用户通过多访问无线电传输将部分任务位卸载到多个MEC服务器，从而实现边缘计算的协作。通过利用非正交多址（NOMA）的优点来提高传输效率，可以期望借助NOMA有效降低多址MEC中用户的能耗。但是，考虑到NOMA的同信道干扰和MEC服务器之间的计算协作，联合任务卸载和资源分配是一个具有挑战性的问题。在本文中，为了最大程度地减少用户的加权总和能量，我们首先针对特殊的单用户（OTORA-SU）和一般多用户（OTORA-MU）情况提出了最佳的任务卸载和资源分配算法，通过利用所提出问题的凸层结构。有趣的是，发现单个用户总是优先将任务位分流给具有更好信道增益的MEC服务器，而不管每个MEC服务器的计算能力如何。然后，针对一般的多用户情况，提出了一种低复杂度的算法（LTORA-MU），该算法收敛速度快，性能接近最佳。考虑到信道估计误差，还研究了信道状态信息不完善（CSI）和连续干扰消除（SIC）的影响。仿真结果表明，对于完美和不完美的CSI和SIC，1）提出的OTORA-SU和LTORA-MU优于本地计算，完全卸载，基于FDMA的卸载和非协作式MEC方案；2）随着MEC服务器数量的增加，