Space-Integrated-Ground Information Networks ›› 2022, Vol. 3 ›› Issue (3): 65-71.doi: 10.11959/j.issn.2096-8930.2022033

Special Issue: 专题:天地融合软件定义网络

• Special Issue: World-Earth Integration Software-Defined Network • Previous Articles     Next Articles

Research on Routing Optimization in Satellite Internet Based on Deep Reinforcement Learning

Linhui WEI1, Guowen LIU1, Yu LIU1,2, Yumei WANG1   

  1. 1 School of Artifi cial Intelligence, Beijing University of Posts and Telecommunications, Beijing 100876, China
    2 Peng Cheng Laboratory, Shenzhen 518000, China
  • Revised:2022-07-24 Online:2022-09-20 Published:2022-09-01
  • Supported by:
    National Key Research and Development Program of China(2019YFB1803103);BUPT Excellent Ph.D.Students Foundation(CX2021113)

Abstract:

With the rapid development of satellite communication, the satellite internet is one of the core technologies of 6G network to realize global coverage, full-time access and full scene service.The high dynamics and limited capacity of satellite network lead to a series of management and control challenges such as heterogeneous network management, dynamic resource allocation and so on.Since the machine learning-based technologies have strength in network design, the intelligent architecture of software-defi ned satellite internet was put forward.In view of the intelligent routing in satellite internet, and leverages the deep reinforcement algorithm based on double delayed deep deterministic policy gradient (TD3) to solve the network routing optimization problem.The experimental results showed that compared with DDPG algorithm, the TD3 algorithm reduced the delay by 19.19%.

Key words: satellite internet, software-defi ned networking, deep reinforcement learning, routing optimization

CLC Number: 

No Suggested Reading articles found!