Research On Edge-based Resource Allocation Algorithm In Metaverse

The concept of the metaverse first appeared in Neil Stephenson’s 1992 science fiction novel Snow Crash.In this novel,Stephenson defines the metaverse as a vast virtual environment parallel to the physical world in which users interact through digital avatars.Today,the Metaverse is considered to be a virtual world that is linked and created by means of technology,and mirrors and interacts with the real world.The metaverse is interconnected,persistent,and massively scalable.First,this paper reviews the development history of the metaverse concept,and analyzes the bottlenecks faced by the development of metaverse-related core technologies.Aiming at the shape of the future metaverse,this paper proposes a layered and clear metaverse hierarchy,which mainly includes the infrastructure layer,the metaverse service layer,and the applications in the metaverse composed of social relations,economic relations,privacy,trust,etc.layer.Considering the requirements for latency,bandwidth,and experience of future metaverse services,this paper proposes an edge-based metaverse service delivery model.In this delivery model,user requirements are offloaded to a specific edge node through an intelligent algorithm,and the edge node loads and renders resources,and finally returns the video stream to the user.In order to speed up the construction of metaverse scenes by nodes,this paper proposes to divide the metaverse according to the granularity of planets,scenes and resources,and organize the rendering nodes into a P2 P network.Secondly,for the resource allocation problem of edge nodes in the metaverse delivery model,this paper considers two methods of static and dynamic resource allocation.Static allocation can be understood as collecting user needs for a period of time for unified allocation.In this paper,the static resource allocation problem is modeled as a multi-objective optimization problem that comprehensively considers user delay and operator service cost.Solve.This paper first introduces the operators of MEGA algorithm,including crossover operator,mutation operator,selection operator,etc.Then,through experiments such as verifying the convergence of the MEGA algorithm,comparing the performance of each algorithm under different problem scales,and comparing the optimization sub-goals of each algorithm,it is proved that the MEGA algorithm has excellent performance,which is about 10%~30% compared to traditional algorithms such as FF,BF,and greedy.effect improved.Although the MEGA algorithm has good performance,it is easily limited by the scale of the problem and the solution time in scenarios with high dynamics and fast decision-making.Aiming at the resource allocation at the edge of the metaverse in dynamic scenarios,this paper models the problem as a multi-objective optimization problem that maximizes user Qo E under resource constraints.The main influencing factors of user Qo E include video stream quality,user delay,etc.Experiments show that the online decision-making scheme based on reinforcement learning has a certain improvement in effect and efficiency compared with traditional algorithms.Under the same problem scale,comparing the online FF and online BF algorithms,prediction through the reinforcement learning model can improve the performance by about 10%~40%.Therefore,the main contributions of this paper are as follows:1.proposes a layered metaverse model and elaborates on the role of each layer.Based on5 G and edge computing technology,an edge-based metaverse service delivery model is proposed.2.Aiming at the static resource allocation problem in the metaverse service delivery model,it is modeled as a multi-objective optimization problem considering operator costs and user delays,and is solved by an improved genetic algorithm MEGA.Compared with the traditional algorithm,MEGA has an effect improvement of about10%~30%.3.Aiming at the dynamic resource allocation problem in the metaverse service delivery model,it is modeled as a multi-objective optimization problem that maximizes user Qo E under resource constraints,and uses reinforcement learning to train the model for decision-making.Compared with the traditional online placement algorithm,there is an improvement of about 10%~40%.