4.7 Article

UAV-Net plus : Effective and Energy-Efficient UAV Network Deployment for Extending Cell Tower Coverage With Dynamic Demands

期刊

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
卷 72, 期 1, 页码 973-985

出版社

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TVT.2022.3202141

关键词

Signal to noise ratio; Throughput; Three-dimensional displays; Solid modeling; Ray tracing; Heating systems; Vehicle dynamics; Coverage path planning; network infrastructure mobility; placement optimization; unmanned aerial vehicles

向作者/读者索取更多资源

Nowadays, in scenarios with insufficient or unstable network bandwidth, unmanned aerial vehicle mounted base stations (UAV-BSs) provide a promising solution. It is important to investigate the effective deployment of UAVs to maximize the sum throughput and service time of mobile clients. This study proposes algorithms based on deep reinforcement learning (DRL) to efficiently solve the NP-hard sub-problems of chunk selection and chunk search, achieving a significant throughput gain with minimal time cost.
Nowadays, mobile data traffic is growing explosively, but in many realistic scenarios, users still often encounter insufficient or unstable network bandwidth. A promising solution is provided by unmanned aerial vehicle mounted base stations (UAV-BSs) to serve regions with bandwidth shortfall. It is significant to investigate how to deploy UAVs effectively and efficiently for maximizing the sum throughput and service time of a set of mobile clients scattered in a large environment. A basic idea is to use RF ray tracing simulations as a hint to narrow down the search space of UAVs for conducting actual measurements. Furthermore, we formulate two key sub-problems, chunk selection, which selects an optimal subset of chunks in the region as the search space of UAVs, and chunk search, which plans the paths of UAVs to cover all selected chunks for conducting measurements, provide network services and charge timely under energy constraints to maximize the total service time. As they are both proved to be NP-hard, a 3D convolutional deep reinforcement learning (DRL) based chunk selection algorithm and an energy-aware DRL-based chunk search algorithm are proposed to solve them efficiently. A prototype system, UAV-Net+, is implemented to evaluate the feasibility by conducting measurements by an UAV mounted WiFi AP communicating with several clients scattered in a campus, reporting an obvious throughput gain with a small measurement overhead and time consumption. Extensive simulations demonstrate the effectiveness of 3D convolutional DRL-based chunk selection algorithm and energy-aware DRL-based chunk search algorithm. It is able to achieve 95.35% effective throughput of the skyline algorithm, Oracle, with only 0.7% time cost. It is more beneficial for providing long-time network service and balancing the workload among UAVs.

作者

我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。

评论

主要评分

4.7
评分不足

次要评分

新颖性
-
重要性
-
科学严谨性
-
评价这篇论文

推荐

暂无数据
暂无数据