DQN Aided Edge Computing in Satellite-Terrestrial Network

In order to support a mass of current satellite applications, it becomes a trend to integrate satellite networks with terrestrial networks, called satellite-terrestrial networks. However, traditional network protocols cannot adapt to the dynamic and complex satellite-terrestrial network. Moreover, the computing and communication capabilities of some satellites cannot meet the requirements of supporting various applications. As a result, the paper proposes an edge computing based software-defined satellite-terrestrial network architecture, which can manage network flexibly by logically centralizing network intelligence and control. Furthermore, a networking and edge computing scheme is proposed by formulating a jointly optimization problem, which is solved by using novel deep Q-learning approach. Simulation results show the effectiveness of the proposed scheme.