移动通信2024，Vol.48Issue(3)：83-89,7.DOI:10.3969/j.issn.1006-1010.20240228-0002
动态拓扑下的无人机网络计算任务卸载方法
Computation Task Offloading for Unmanned Aerial Vehicle-Enabled Networks with Dynamic Topology
摘要
Abstract
Addressing the computation task offloading challenge in unmanned aerial vehicle(UAV)-enabled networks with dynamic topology,this paper introduces a high-performance,low-complexity computation task offloading approach leveraging an attention mechanism and deep reinforcement learning.The attention mechanism dynamically represents the number of UAVs in the network,overcoming the limitation of traditional deep reinforcement learning-based offloading methods that are tailored only for static network topologies.Building upon this,a cascaded training methodology combining pre-training with reinforcement learning is proposed,significantly enhancing the convergence speed and performance of the method.Simulation results demonstrate that the proposed algorithm substantially reduces the system's average latency and packet loss rate compared to benchmark schemes.关键词
无人机通信/计算任务卸载/注意力机制/强化学习Key words
unmanned aerial vehicle communication/computation task offloading/attention mechanism/deep reinforcement learning分类
信息技术与安全科学引用本文复制引用
职科翔,李刘杰,刘晨熙,李长庚,彭木根..动态拓扑下的无人机网络计算任务卸载方法[J].移动通信,2024,48(3):83-89,7.基金项目
国家重点研发计划项目"6G通信-感知-计算融合网络架构及关键技术"(2021YFB2900200) （2021YFB2900200）
中国通信学会青年人才托举项目(YESS20200064) （YESS20200064）
