Software-defined satellite-ground convergence network QoE perception routing architecture based on deep reinforcement learning

Abstract

The invention discloses a software-defined satellite-ground convergence network QoE (Quality of Experience) sensing routing architecture based on deep reinforcement learning, which utilizes the centralized control and programmable capabilities of a software-defined satellite-ground convergence network, organically combines the self-learning capability of deep reinforcement learning, and takes user QoE as a routing planning basis. A super controller, a domain controller and a slave controller which are physically distributed are adopted to realize collection and updating of network states, calculation of inter-domain routing, mapping and convergence of service flows, distribution of routing instructions and configuration of satellite switch flow table entries; and deploying a multimedia service QoE evaluation module and a deep reinforcement learning agent on the intra-domain controller, and calculating an intra-domain forwarding path by using a deep reinforcement learning algorithm by taking a user QoE value fed back by the QoE evaluation module as an award. According to the method, the calculation load between the controllers can be balanced, the accuracy requirements on the calculation capability of the controllers and the dynamics analysis and service flow modeling of the satellite-ground fusion network are reduced, and the user experience quality of the whole network is improved.

Description

technical field [0001] The invention relates to the field of satellite-ground fusion network communication and routing, in particular to a software-defined satellite-ground fusion network QoE-aware routing architecture based on deep reinforcement learning. Background technique [0002] The 6G network will break through the limitations of the terrain and surface, truly realize the three-dimensional coverage of global users' full-scene information no matter when and where, respond to the future demand for intelligent communication of all things, and provide users with the ultimate performance experience in more complex and diverse application scenarios. Building a cross-regional, cross-airspace, and cross-sea satellite-ground fusion network to achieve truly seamless global coverage is an important direction for the development of future communication networks. Aiming at the problems of poor flexibility, weak differentiated services, and low resource utilization of the satellit...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to face the routing planning problem of the software-defined satellite-ground fusion network. Around the satellite-ground fusion network routing method, it is assumed that the network environment and user needs can be accurately modeled, and the network QoS parameters are used as the routing planning basis, resulting in the user QoE can not be achieved. It is well satisfied, and at the same time cannot dynamically adapt to network changes. The present invention takes user QoE as the basis for routing planning, fully utilizes the centralized control and programmable capabilities of the software-defined satellite-ground fusion network, and combines the depth of self-learning capabilities of deep reinforcement learning. Combined, a QoE-aware routing architecture based on deep reinforcement learning for software-defined satellite-ground fusion network is proposed

Images