科研之友
微信
新浪微博
Facebook
分享链接
摘要
The communication network of 5G is a heterogeneous network that integrates multiple access technologies. It supports a large number of diversified services from different vertical industries, including high definition (HD) videos, augmented reality, and autonomous driving. It is difficult for traditional networks to tackle this situation via QoS policies. Therefore, network slicing has become an important technology for 5G networks. At present, research on network slicing focuses on 5G core network slicing, while wireless network slicing is still burgeoning due to its certain complexity and technical difficulty. In this paper, we propose a slice construction method for wide-narrow fusion networks. First, based on the QoS requirements of different service, service flow information, and network topology information, we construct the slice paths and the mapping of network resources. Second, we develop a network slice scheduling algorithm based on the deep reinforcement learning model for wide-narrow fusion networks. We take the real-time network performance information of the constructed slice and the QoS requirements of the service as the state parameters for the neural network, and train the model to obtain the optimal scheduling strategy of the slice for different network environments. The results of comparative simulation experiments show that the proposed algorithm achieve a great improvement in the QoS satisfaction rate of the service, the average utilization rate of slice bandwidth, and the completion time of service scheduling. The algorithm can enable the customized transmission of service data in heterogeneous networks, and help to efficiently and flexibly control wideband- narrow-band integrated networks. ? 2022 Journal of Nanjing Institute of Posts and Telecommunications.
- 单位
