Research On Multi Dimensional Heterogeneous Resources Management For Satellite-Terrestrial Integrated Networks

With more than 50%of the world still lacking high-speed Internet access services,especially in rural and remote areas,satellite communications have an obvious advantage for global coverage,so the effective integration of satellite and terrestrial networks has become an important part of the 6th Generation Mobile Communication Technology(6G)wireless architecture.However,satellite payloads have weak spacebased computing capabilities,but for 6G,the need for computing and cache of services is prevalent,whether in hot spots or sparse areas for communication services,so seamless coverage is still a challenge.Mobile Edge Computing(MEC)is a technology that has received a lot of attention in recent years.The introduction of MEC into satelliteterrestrial integrated networks can provide users with multi-dimensional,heterogeneous computing and cache resources,and space-based computing can enable the streaming and processing of computing tasks and cache content at low-earth-orbit(LEO)satellite nodes,saving bandwidth and reducing the impact of large latency on satellite-terrestrial or inter-satellite links.In the heterogeneous satellite-terrestrial network,edge nodes are connected in a loosely coupled manner,making the coordinated management of cache,computing,and communication resources more complex.Therefore,how to manage the multi-dimensional heterogeneous resources in the satellite-terrestrial integrated networks so that they can be efficiently utilized has become an urgent problem to be solved.To address the communication,computing,and cache issues arising from providing seamless coverage to users in satellite-terrestrial integrated networks,this thesis uses Deep Reinforcement Learning(DRL)to build a computing offloading and edge cache scheme to manage resources and explore the implementation of a satellite-based coverage extension mechanism.The research of this thesis focuses on three aspects.Firstly,for the problems of weak coverage and no coverage in remote areas and hotspot areas,a wide area coverage expansion mechanism based on GEO and LEO satellite collaboration is proposed to obtain communication equipment spatiotemporal information and service demand through joint coverage of GEO and LEO satellite networks,obtain the change pattern and distribution law of key information of various coverage situations.GEO satellite sensing prediction and LEO satellite coverage enhancement,achieving wide coverage in remote areas and deep coverage enhancement in hotspot areas on satellite access.Secondly,a LEO satellite-terrestrial collaborative edge computing(LSTEC)network with a three-tier computing architecture is designed,where the user can choose to execute the task locally,or partially offload the task to the base station edge server the LEO satellite edge server for remote execution.Considering the difference in computing power between the LEO satellite and the base station,as well as the geometric model of the satellite coverage users,it is modeled as a Nonlinear Programming problem for jointly optimizing the task offloading ratio decision and the allocation of computing and communication resources.A deep deterministic policy gradient(DDPG)-based task offloading algorithm is proposed and validated with simulation results showing that the proposed algorithm outperforms existing schemes in terms of latency and energy consumption.Finally,a multi-level collaborative edge cache network architecture is proposed for the optimization of edge cache placement in satelliteterrestrial integrated networks,where users can obtain the requested content at the base station,LEO satellite,and cloud server.A user preference-driven and hybrid popularity-based cache strategy are designed,and algorithms for user association,cache,and communication resource allocation based on DDPG are proposed.The proposed algorithm is evaluated based on the empirical Zipf distribution and TMDB dataset,and simulation results show that the proposed algorithm outperforms traditional studies in terms of reduced latency and improved cache hit ratio.