| 引用本文： |
-
黄启明,袁正林,龚正伟,等.一种基于DQN的卫星通信车载站中频功率智能控制方法[J].电讯技术，2025，65（7）：1120 - 1128. [点击复制]
- HUANG Qiming,YUAN Zhenglin,GONG Zhengwei,et al.An Intelligent Intermediate-frequency Power Control Method for Satellite Communication Vehicle Stations Based on DQN[J].，2025，65（7）：1120 - 1128. [点击复制]
|
|
| 摘要: |
| 针对传统卫星通信车载站中频信号功率控制策略中存在的自动化程度低、控制效率低下等问题，提出了一种基于深度Q学习网络（Deep Q-learning Network,DQN）的功率智能控制方法。将功率控制决策转化成一个马尔可夫决策过程：选取信道终端设备（Channel Terminal，CT）参数构建状态空间，以终端链路操作和禁呼时间构建动作空间，设计了基于业务价值的奖励函数和基于物理特性的状态转移函数。提出的控制策略实现了中频信号功率控制智能化，算法收敛状态平均回报可以达到主流深度强化学习算法水平，平均回合训练时间仅为对照组最长时间的6.45%。 |
| 关键词: 卫星通信车载站 中频功率控制 深度Q学习网络(DQN) |
| DOI：10.20079/j.issn.1001-893x.240729004 |
|
| 基金项目: |
|
| An Intelligent Intermediate-frequency Power Control Method for Satellite Communication Vehicle Stations Based on DQN |
| HUANG Qiming,YUAN Zhenglin,GONG Zhengwei,SONG Jun |
| (1.College of Information Science and Technology/College of Artificial Intelligence,Nanjing Forestry University,Nanjing 210037,China;2.CETC Advanced Mobile Communication Innovation Center,Shanghai 200331,China) |
| Abstract: |
| In response to the issues of low automation and inefficient control in the power control strategies of traditional satellite communication vehicle stations,a power intelligent control method based on deep Q-learning network(DQN) is proposed.The power control decision-making process is formulated as a Markov decision process(MDP):the channel terminal device(CT) parameters are selected to construct the state space,while the action space is defined according to terminal link operations and call blocking times.Additionally,a reward function based on service value and a state transition function based on physical characteristics are designed.The proposed control strategy achieves the intelligent power control of intermediate-frequency signals.The average return in the convergence state of the algorithm can reach the level of mainstream deep reinforcement learning algorithms,with the average training time per episode being only 6.45% of that of the longest duration in the control group. |
| Key words: satellite communication vehicle station intermediate frequency power control deep Q-learning network(DQN) |
《电讯技术》编辑部 