当前位置: X-MOL 学术IEEE Netw. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Intelligent Task Offloading and Energy Allocation in the UAV-Aided Mobile Edge-Cloud Continuum
IEEE NETWORK ( IF 6.8 ) Pub Date : 2021-11-08 , DOI: 10.1109/mnet.010.2100025
Zhipeng Cheng 1 , Zhibin Gao 1 , Minghui Liwang 1 , Lianfen Huang 1 , Xiaojiang Du 2 , Mohsen Guizani 3
Affiliation  

The arrival of big data and the Internet of Things (IoT) era greatly promotes innovative in-network computing techniques, where the edge-cloud continuum becomes a feasible paradigm in handling multi-dimensional resources such as computing, storage, and communication. In this article, an energy constrained unmanned aerial vehicle (UAV)-aided mobile edge-cloud continuum framework is introduced, where the offloaded tasks from ground IoT devices can be cooperatively executed by UAVs acts as an edge server and cloud server connected to a ground base station (GBS), which can be seen as an access point. Specifically, a UAV is powered by the laser beam transmitted from a GBS, and can further charge IoT devices wirelessly. Here, an interesting task offloading and energy allocation problem is investigated by maximizing the long-term reward subject to executed task size and execution delay, under constraints such as energy causality, task causality, and cache causality. A federated deep reinforcement learning (FDRL) framework is proposed to learn the joint task offloading and energy allocation decision while reducing the training cost and preventing privacy leakage of DRL training. Numerical simulations are conducted to verify the effectiveness of our proposed scheme as compared to three baseline schemes.

中文翻译:


无人机辅助移动边缘云连续体中的智能任务卸载和能量分配



大数据和物联网时代的到来极大地促进了网内计算技术的创新,边云连续体成为处理计算、存储、通信等多维度资源的可行范式。在本文中,介绍了一种能量受限无人机(UAV)辅助的移动边缘云连续体框架,其中从地面物联网设备卸载的任务可以由无人机作为边缘服务器和连接到地面的云服务器协同执行基站(GBS),可以看作是接入点。具体来说,无人机由GBS发射的激光束提供动力,并可以进一步为物联网设备进行无线充电。在这里,通过在能量因果关系、任务因果关系和缓存因果关系等约束下最大化受执行任务大小和执行延迟影响的长期奖励来研究一个有趣的任务卸载和能量分配问题。提出了联邦深度强化学习(FDRL)框架来学习联合任务卸载和能量分配决策,同时降低训练成本并防止 DRL 训练的隐私泄露。进行数值模拟以验证我们提出的方案与三个基线方案相比的有效性。
更新日期:2021-11-08
down
wechat
bug