Research On Matching-Based Resource Optimization For Space-Air-Ground Integrated Network

As the rapid development of society and communication technology,human beings have entered the era of ubiquitous mobile interconnection.Existing terrestrial 5G network has limited coverage and insufficient resource allocation,which results in severe challenge for network applications.SAGIN 6G support ubiquitous and reliable access for large numbers of mobile users and IoT terminals at any location.However,current network resource allocation methods mainly focus on system QoS,but rarely take users’ specific needs for network content into consideration.And due to high dynamics of network,both network topology and channel state information(CSI)are time varying,making resource allocation intractable.To cope with the problem of satisfying users’ specific demand for network content,in this paper we propose a resource allocation system that includes three entities:content source,SAGINE and end users.To deal with the issue of SAGIN dynamics,we leverage machine learning to enhance matching,and propose a solution based on matching-learning.Both theoretical analysis and simulations are carried out.To sum up,the main work of this paper is as follows:(1)First,for the problem of users’ specific demand for network content,we propose a system model for resource allocation with three entities of content source,SAGINE and terminal users in SAGIN 6G.In this model,users require to establish a connection with SAGINE,which forwards the requests to content service providers.Since finding the optimal stable matching with the maximum cardinality is an NP-complete problem,we design a decoupled resource allocation algorithm,which separates the three types of entities into two,i.e.,content source-SAGINE pairs and users.In this way,the matching problem of three entities is transformed into a two-sided matching issue,which is solved with low complexity.(2)Second,since two-sided matching approach needs to decouple the relationship among three entities,it causes the lack of preference information for each entity and results in system performance loss.Thus,we reformulate the problem as a Restricted Three-Sided Matching with Size and Cyclic(R-TMSC)by adding some reasonable constraints.Then,the content-oriented R-TMSC resource allocation(COR2A)algorithm and the user-oriented R-TMSC resource allocation(UOR2A)algorithm are developed to solve the above issue.Simulation results show that the three-sided matching algorithm can increase the CSP revenue as well as the number of users,and improve user satisfaction.(3)Finally,due to high dynamics of SAGIN,it is difficult for a user to obtain the global knowledge.In case of information uncertainty,we enhance matching with machine learning and provide the CVA-UCB solution based on MAB,where end users unceasingly attempt and learn the optimal access solution by successively observing the relevance among its preference to UAVs,matching triples,and transmission rate performance.The simulation results indicate that the proposed method can effectively balance exploration and exploitation during the learning process,and significantly improve the data transmission rate.