Research On Task Offloading And Resource Management Technology Of Multi-Beam Satellite Network

Although terrestrial communication technology can meet most of the broadband demand in urban areas,the broadband demand in remote areas strongly depends on satellite services.Recently,the increasing number of satellite communication terminals and the emergence of emerging multimedia applications have caused a large-scale increase in communication demand,which has made satellite suppliers switch from the traditional single-beam architecture to a high-throughput satellite system with multiple spot beams,which reuses carrier bandwidth on multiple beams to meet the growing communication demand.However,the status quo of limited satellite communication network resources has not changed,so it needs to be reasonably utilized to avoid waste and meet the required user performance.The technology of Mobile Edge Computing is integrated into the satellite communication system,which can perform task computing or data storage at the edge,reduce the amount of data that must be returned through the satellite link,and in some cases,reduce the data transmission latency.In the satellite communication system,the spatial distribution of terminals and their business requirements is uneven,and some satellite resources covering high-demand areas are insufficient,while some satellite resources covering low-demand areas are surplus and the resource utilization rate is low.Moreover,the distribution of terminals and their requirements is time-varying,so it is difficult for satellite edge nodes with limited resources to meet the needs of large-scale terminal services during peak hours.On the other hand,with the increasing number of Internet of Things terminals,the problem of resource allocation becomes more complicated,and the computational complexity of the corresponding resource allocation algorithm is also higher.Therefore,this paper will study the problem of low resource utilization caused by uneven distribution of terminals,the problem that satellite edge nodes with limited resources cannot meet the needs of large-scale terminal services during peak hours,and the problem of high computational complexity of traditional resource allocation algorithms in large-scale terminal scenarios.Firstly,in order to solve the problem of low resource utilization caused by uneven distribution of terminals,this paper studies the hybrid GEO-LEO satellite Internet of Things network with joint computing and communication resource allocation.In order to solve the problem of load imbalance,the tasks generated by IoT devices can be processed by LEO satellites with cooperative characteristics or forwarded to the ground gateway station through GEO satellites for processing.The joint task offloading,communication and computing resource allocation of hybrid satellite Internet of Things network can be expressed as a mixed integer dynamic programming problem with satellite-ground cooperation and inter-satellite cooperation through inter-satellite links.In order to solve this problem,this paper proposes an intelligent task offloading and multidimensional resource allocation algorithm(TOMRA),which uses a method based on deep reinforcement learning and convex optimization to solve the sub-problems of task offloading and channel allocation and the sub-problems of computing resource allocation under fixed offloading and channel allocation decisions,respectively,so as to minimize the latency of task offloading and processing.Through simulation analysis,it is verified that this algorithm can better schedule resources and offload tasks according to the load distribution characteristics compared with other comparative algorithms,so as to minimize the system latency.Secondly,aiming at the problem of high complexity of resource allocation algorithm,considering the distribution characteristics of service requirements in satellite communication network and the relative stability of beam-level service requirements and the real-time characteristics of user-level service requirements,this paper proposes a beam-level and userlevel hierarchical resource management architecture based on double time scales,in which beam-level and user-level schedule and allocate resources in different time scales.After beam-level resource allocation is completed,user-level resource allocation is carried out independently and in parallel in each beam,thus reducing the action space of resource allocation and reducing the complexity of resource allocation problem.On this basis,a beam-level and user-level hierarchical resource allocation algorithm is proposed for large-scale end-user data download scenarios,and the algorithm has achieved convergence after three stages of training.Finally,the algorithm and user-level resource allocation algorithm are compared and analyzed by simulation,which verifies the effectiveness of the beamlevel and user-level hierarchical resource management architecture and the hierarchical resource allocation algorithm based on double agents.