中国邮电高校学报(英文版) ›› 2019, Vol. 26 ›› Issue (3): 35-43.doi: 10.19682/j.cnki.1005-8885.2019.0016

• Wireless • 上一篇    下一篇

Dynamic power control for relay-aided transmission based on deep reinforcement learning

秦彩1,王朝炜1,王卫东2,张英海1   

  1. 1. 北京邮电大学
    2. 北京邮电大学电子工程学院
  • 收稿日期:2019-01-25 修回日期:2019-04-09 出版日期:2019-06-30 发布日期:2019-06-30
  • 通讯作者: 王朝炜 E-mail:wangchaowei@bupt.edu.cn
  • 基金资助:
    中国国家重点研发计划;北京市人才培养

Dynamic power control for relay-aided transmission based on deep reinforcement learning

  • Received:2019-01-25 Revised:2019-04-09 Online:2019-06-30 Published:2019-06-30
  • Supported by:
    National Key R&D Program of China

摘要: Using relay in the wireless communication network is an efficient way to ensure the data transmission to the distant receiver. In this paper, a dynamic power control approach (DPC) is proposed for the amplify-and-forward (AF) relay-aided downlink transmission scenario based on deep reinforcement learning (DRL) to reduce the co-channel interference caused by spectrum sharing among different nodes. The relay works in a two-way half-duplex (HD) mode. Specifically, the power control of the relay is modeled as a Markov decision process (MDP) and the sum rate maximization of the network is formulated as a DRL problem. Simulation results indicate that the proposed method can significantly improve the system sum rate.

关键词: power control, deep reinforcement learning, relay, downlink

Abstract: Using relay in the wireless communication network is an efficient way to ensure the data transmission to the distant receiver. In this paper, a dynamic power control approach (DPC) is proposed for the amplify-and-forward (AF) relay-aided downlink transmission scenario based on deep reinforcement learning (DRL) to reduce the co-channel interference caused by spectrum sharing among different nodes. The relay works in a two-way half-duplex (HD) mode. Specifically, the power control of the relay is modeled as a Markov decision process (MDP) and the sum rate maximization of the network is formulated as a DRL problem. Simulation results indicate that the proposed method can significantly improve the system sum rate.

Key words: power control, deep reinforcement learning, relay, downlink