Research On Joint Communication And Computation Scheduling Methods In Integrated Satellite-terrestrial Networks

Satellite communication covers a large area and is less affected by terrain.It can flexibly support access services in a wide range of areas,and can provide capacity supplements in densely populated areas and range supplements in remote areas for terrestrial networks.Therefore,the concept of integrated satellite-terrestrial networks emerged.In recent years,with the booming development of emerging Internet applications,the service number of the integrated satellite-terrestrial network has increased,and the service type has expanded from basic data transmission to computation services.Hence,countries around the world have started building ultra-dense Low Earth Orbit(LEO)integrated satellite-terrestrial networks to enhance the network service capability.The ultra-dense LEO integrated satellite-terrestrial network provides more communication and computation resources by deploying a large number of satellites.However,a large number of resources also make the scheduling of communication and computation more complicated,especially considering high dynamic network environments.A series of studies are conducted to address the above issues.First,in order to solve the co-frequency interference problem in ultra-dense LEO integrated satellite-terrestrial networks,the communication resource scheduling method in data transmission scenarios is studied,which provides a basis for subsequent research.Second,based on the location of computing resources,the way of supplying resources,and the service requirements,the joint communication/computation resource and task scheduling methods in ground-and satellite-based computation offloading scenarios are investigated to provide better computation services.Finally,the research results are as follows.1.We designed a multi-band switching data transmission strategy in ultra-dense LEO integrated satellite-terrestrial networks.Aiming at the problem of co-channel interference and the differentiated characteristics of different bands affected by the environment,an ultradense LEO integrated satellite-terrestrial network architecture with variable bands is proposed,which can reduce co-channel interference and adapt to dynamic environments by flexibly satellite band switching.Considering the demand of high data transmission rate and the cost of satellite band switching,a two-timescale joint optimization problem of satellite access and satellite band switching is established in a data transmission scenario.This scheme makes satellite access decisions in small-timescale and satellite band switching decisions in large-timescale,thus increasing the data rate while avoiding frequent satellite band switching.2.We designed a ground-based computation offloading strategy based on dueling double deep Q network in ultra-dense LEO integrated satellite-terrestrial networks.In the scenario where computing resources are deployed on the ground station,the computation tasks of ground users are offloaded to the ground station through LEO satellites.The influence of changing uplink and downlink channel conditions and the backlog status of the task queue of the ground station on the total delay needs to be considered simultaneously.Therefore,based on the differential movement rates of different satellites,weather conditions,and queue status of the users and ground stations,a joint optimization strategy of task scheduling and resource allocation is designed,aiming to improve the computation tasks’ quality of service(QoS)and reduce satellite power consumption through reasonable user-satellite-ground station offloading path selection and resource scheduling.In addition,aiming at the difficulty of delay statistics due to the change of satellite-ground connection during task transmission,the correlation of continuous tasks is used to convert task-grained delay statistics into queue-grained delay statistics,which reduces the complexity of the problem.3.We designed a satellite-based computation offloading strategy based on Stackelberg game in ultra-dense LEO integrated satellite-terrestrial network.For the scenario where computation resources are deployed on satellites,the satellite operator provides on-satellite computation resources and charges for them,and ground users offload computation tasks and pay for them accordingly.Therefore,an incentive mechanism is needed to enable the satellite operator and terrestrial users to make reasonable pricing and task offload scheduling decisions to simultaneously increase their utility.A Stackelberg game model for satellite-based computation offloading is proposed to describe the incentive mechanism between a satellite operator and terrestrial users,and analyzes the Stackelberg equilibrium.However,in a highly dynamic ultra-dense LEO integrated satellite-terrestrial network,the traditional optimization algorithm needs to be repeated every time the environment changes.Moreover,due to privacy considerations,ground users are reluctant to share decision-making information with other users.Consequently,a distributed reinforcement learning-based algorithm is proposed for satellite operator’s pricing and ground users’ computation offloading.By training and using the learning model,each node can generate equilibrium solutions independently and quickly,and obtain high utility for satellite operators and ground users.