| 摘要: |
| 为满足空天网络中超可靠低时延通信需求，构建了低轨（Low Earth Orbit，LEO）卫星移动性模型和空天网络中舱内与舱外下行通信模型，在带宽敏感用户的容量约束和飞机的数据积压约束下，研究了最小化时延敏感用户平均传输时延的功率与信道资源联合分配问题。针对该动态资源分配问题，提出了一种基于深度Q网络（Deep Q-network，DQN）的算法。该方法将强化学习与神经网络相结合，把LEO卫星子信道状态和舱内接入点的信道状态等网络状态作为状态空间，LEO卫星子信道的连接关系、LEO卫星的子信道功率分配和舱载用户的功率分配方案作为动作空间，时延敏感用户的平均传输时延作为奖励函数，根据动态环境自适应调整最优分配方案。仿真结果表明，在空天通信动态环境中，与传统算法相比，所提算法可降低15.4%的传输时延。 |
| 关键词: 低轨卫星通信 空天通信 高动态网络 动态资源分配 深度Q网络 |
| DOI：10.20079/j.issn.1001-893x.240925004 |
|
| 基金项目:国家自然科学基金资助项目(62201274，62201275) |
|
| Dynamic Resource Allocation Algorithm in Space-Air Networks |
| SHI Jianfeng,LU Chen,LI Baolong |
| (School of Electronics and Information Science Engineering，Nanjing University of Information Science and Technology，Nanjing 210044，China) |
| Abstract: |
| To meet the requirement of ultra-reliable low-latency communication in space-air networks，a low Earth orbit(LEO) satellite mobility model and a downlink communication model for in-cabin and out-of-cabin scenarios are constructed.Under the constraints of bandwidth-sensitive user’s capacity and data backlog，a joint allocation problem of power and channel resources is studied to minimize the average transmission latency for latency-sensitive users.A deep Q-network(DQN) algorithm is proposed to address the dynamic resource allocation problem.This method combines reinforcement learning with neural networks，using the network states such as the subchannel states of the LEO satellite and the channel states of onboard access points as the state space.The connection relationships of LEO satellite subchannels，the subchannel power allocation of the LEO satellite，and the power allocation schemes for onboard users are taken as the action space.The average transmission latency of delay-sensitive users is the reward function，enabling adaptive adjustment of the optimal allocation scheme according to the dynamic environment.The simulation results show that in the dynamic environment of space-air communication，the proposed algorithm，compared with traditional algorithms，can reduce transmission latency by 15.4%. |
| Key words: LEO satellite communication space-air communication high dynamic network dynamic resource allocation deep Q-network |
《电讯技术》编辑部 