Multihop Task Routing in UAV-Assisted Mobile-Edge Computing IoT Networks With Intelligent Reflective Surfaces
Multihop Task Routing in UAV-Assisted Mobile-Edge Computing IoT Networks With Intelligent Reflective Surfaces
复制标题
DOI：
10.1109/jiot.2022.3228863
复制
发表时间：
2023-04
影响因子：
10.6
通讯作者：
Yousef N. Shnaiwer;N. Kouzayha;M. Masood;Megumi Kaneko;T. Al-Naffouri
中科院分区：
文献类型：
--
作者：
Yousef N. Shnaiwer;N. Kouzayha;M. Masood;Megumi Kaneko;T. Al-Naffouri
无人机（uav）与地面移动边缘计算（MEC）服务器在处理任务方面的协作正成为MEC网络的主要研究方向之一。尽管无人机辅助MEC具有优势，但受到无人机电池容量有限和能耗敏感的制约。与之前允许无人机在本地处理任务或将其卸载到地面MEC服务器的工作不同，在本文中，我们提出了一种用于物联网（IoT）网络的多跳任务路由解决方案，其中无人机还可以中继到另一架与地面MEC服务器连接更好的无人机。此外，无人机可以利用现有的智能反射面（IRSs）进一步改善任务卸载和降低能耗。我们证明了最小化无人机总能量的问题是np困难的，并提出了一种基于图的启发式解决方案。仿真结果表明，该方案优于传统的无无人机-无人机中继方案，特别是在部署irs时。在此基础上，设计了卷积神经网络（CNN）来减少无人机在中心化协调器上寻找决策的延迟。仿真表明，与基于图的启发式解决方案相比，CNN实现了非常接近的能耗性能和显着减少的执行时间。
The cooperation between unmanned aerial vehicles (UAVs) and ground mobile-edge computing (MEC) servers in processing tasks is becoming one of the main research trends of MEC networks. Despite the advantages of UAV-assisted MEC, it is restricted by the limited battery capacity and sensitive energy consumption of UAVs. Unlike the previous works where UAVs are allowed to either process tasks locally or offload them to ground MEC servers, in this article, we propose a multihop task routing solution for Internet of Things (IoT) networks in which a UAV can also relay to another UAV with better connection to a ground MEC server. Furthermore, the UAV can make benefit of existing intelligent reflective surfaces (IRSs) to further improve task offloading and reduce energy consumption. We show that the problem of minimizing the total energy of UAVs is NP-hard, and we propose a graph-based heuristic solution to solve it. Simulation results show that the proposed graph-based solution outperforms the traditional no UAV–UAV relaying scheme, especially when IRSs are deployed. Furthermore, a convolutional neural network (CNN) is devised to reduce the delay of finding the decisions for the UAVs at the centralized coordinator. Simulations show that the CNN achieves very close energy consumption performance and a remarkable reduction in execution time compared to the graph-based heuristic solution.