Machine Learning for Space–Air–Ground Integrated Network Assisted Vehicular Network: A Novel Network Architecture for Vehicles
Fengxiao Tang, Cong Wen, Ming Zhao, Nei Kato
Abstract
The future intelligent network requires seamless coverage, low latency, and ultrareliable communication, which is far from enough for the current vehicular network. The space–air–ground integrated network (SAGIN) includes the ground, air, and space network segments, which is considered a new network architecture for future intelligent communications. Then, we introduce a novel network architecture that uses SAGIN to assist the vehicular network to form a space–air–ground integrated vehicular network (SAGIVN). By using a SAGIN assist vehicular network, the vehicular network can obtain unprecedented performance. However, due to the highly heterogeneous and dynamic characteristics of SAGIVN, the management of resource pools and the conversion of communication among heterogeneous nodes are considerable challenges. As a powerful artificial intelligence tool, machine learning plays a vital role in wireless networks. In this article, we introduce several applications of machine learning technology in SAGIVN. In order to better illustrate the feasibility of the application of machine learning in SAGIVN, we further propose a case study of a traffic offloading scheme based on federated reinforcement learning in SAGIVN.