Research On Resource Allocation Technology Based On Machine Learning For Optical And Wireless Converged Networks

Optical and wireless converged network has become an important networking form of mobile communication system.Emerging applications in the next generation of mobile communication,such as smart city,autonomous driving and telemedicine,have brought the tide of terminal access traffic,and put forward higher requirements on the delay and bandwidth of network services.In order to effectively integrate various emerging applications and ensure the quality of service while meeting their flexible and changeable demands,optical and wireless converged network is in urgent need of more efficient,reasonable and intelligent resource allocation schemes to make full use of limited network resources.In optical and wireless converged networks,the limitations of traditional resource allocation techniques are as follows:(1)the service characteristics and network resource status cannot be accurately perceived;(2)the access trafic tide cannot be accurately predicted and targeted resource reservation cannot be made;(3)the resource allocation scheme has high computational complexity and is easy to fall into local optimal solution.Therefore,starting from machine learning technology,this thesis focuses on the intelligent resource allocation technology of optical and wireless converged network based on network resource state perception and access traffic tide prediction,and has achieved innovative results.The main contents include:Firstly,in the face of the problem of low accuracy of network resource state perception in optical and wireless converged network,this thesis designs a resource perception method based on integrated bidirectional circulating neural network.By constructing a composite integration model containing multiple bidirectional cyclic neural networks,the multi-dimensional,long-term and short-term network traffic characteristics are extracted with different time windows,and the occupation of network resources is studied,so as to realize the real-time and accurate perception of the available state of network resources.Secondly,in the face of the low accuracy of traffic tide prediction in current optical and wireless converged network,this thesis studies the edge cloud collaborative traffic prediction mechanism for optical and wireless converged network nodes.By comprehensively utilizing the network location and resource advantages of cloud platform and edge nodes,the edge cloud collaboration mechanism is designed to form a global perception and prediction that considers multiple perspectives,which improves the traffic prediction accuracy of system.Thirdly,in the face of the current optical and wireless converged network is difficult to timely deal with the problem of sudden traffic peak,this thesis focuses on the intelligent resource allocation strategy of optical and wireless converged network based on traffic prediction.A dynamic power resource planning algorithm based on convex optimization was proposed,and the resource allocation was expressed as a convex optimization problem.The solution space of global optimal solution was reduced by using traffic prediction results,and the solving efficiency of optimal resource allocation scheme was improved.A network resource allocation strategy for sudden traffic tide is designed.By defining resource evaluation factor and service volume scaling factor,intelligent resource scheduling for sudden traffic peak is realized,which effectively improves the network’s capacity of carrying and dredging traffic tide.Simulation results show that the proposed resource allocation strategy can make reasonable resource reservation according to traffic prediction results,and achieve high throughput while reducing power cost.It can effectively reduce blocking rate and improve resource utilization rate when dealing with sudden tidal traffic.