Research On Intelligent Control Technology Of Converged Optical-Wireless Network

With the rapid development of 5G/B5G,in order to meet the ultra-low latency requirements of emerging applications such as the Internet of Vehicles,augmented reality,and industrial Internet,the access network architecture of convergenced optical-wireless network has become an important component of 5G/B5G.Due to the dynamics and differentiation of services in the 5G/B5G era,facing different network requirements,the network needs end-to-end slicing to cope with differentiated services.Converged Optical-wireless network slicing configuration relies on the experience of operators to operate the network,which lacks flexibility and timeliness.In the B5G scenario,users are more concerned about the implementation of configuration,rather than the process of configuring operations on the network.In order to effectively accommodate various services in 5G/B5G and meet the requirements of emerging services,new requirements are put forward for the flexibility and scalability of resource management in converged optical-wireless networks.In summary,the implemention of network self-learning,self-optimization and self-management is the future development direction of intelligent control technology for convergenced optical-wireless network.Therefore,from the perspective of intelligent slicing strategy,this thesis focuses on the intelligent control strategy based on intent-driven converged optical-wireless network,and has obtained several innovative research results on the closed-loop control of intent-driven converged optical-wireless network slicing.The main contents are as follows:Firstly,for the existing converged optical-wireless network,it is difficult to accurately and effectively convert the business intent into the configuration language of the optical and wireless converged network.This paper studies the mapping of research intent and network resource requirements,and captures the business intent to build a knowledge graph.Using the knowledge graph matching technology to convert it into a network strategy,realizing the precise conversion from intent to network state requirements.Secondly,in view of the difficulty of the existing converged optical-wireless network to provide customized services for the flexible needs of diversified intents,this paper studies the generation of slicing strategies based on reinforcement learning(DRL)under the constraints of intents.By effectively converting user performance requirements into converged optical-wireless network slicing configuration strategies,to achieve multi-dimensional sensing requirements,and the optimal adaptation of intent and strategy,to overcome the shortcomings of static solidification of traditional converged optical-wireless network resource allocation strategies.Thirdly,when the intelligent slicing strategy is issued and executed,due to the dynamic nature of the intention,it may face the problem of not being able to adapt to the sudden network environment.This paper focuses on the intent guarantee mechanism based on the rapid slicing reconstruction method,by introducing Dueling DQN network proposes a fast slicing reconstruction method for highly dynamic network environments,which effectively adjusts the allocation strategies of computing and storage resources in different domains to match the dynamic changes in slicing requests,and maximizes the realization of intent.The simulation results show that the proposed model can continuously detect the consistency between the execution strategy and the original intention according to the real-time network status,and adopts fast slice reconstruction when the intent requirements are not met.The reconstruction time is shortened by 39.2%,and the resource utilization rate is increased by 17.6%,to reduce the rate of violations of intention constraints while increasing the utilization of network resources.