【作者】 刘宇；

【导师】 寿国础；

【作者基本信息】 北京邮电大学 ， 信息与通信工程， 2025， 博士

【摘要】 随着卫星性能的持续提升,通过将基站和核心网关键网络功能部署至卫星,赋予卫星网络部分核心网功能,从而实现在卫星上直接处理信令和转发数据,避免了信令和数据回传至地面。星载核心网不仅显著降低了信令和数据包的传输时间,还有效缓解了卫星与地面信关站之间的信令风暴,能够在无需地面基础设施的情况下,直接为用户提供服务。然而,地面网络中基础设施固定,且相较于卫星,其资源更为丰富。而卫星高速移动且资源匮乏,使得传统为地面网络设计的核心网难以适应动态变化的卫星网络环境。星载核心网主要面临以下三个挑战:1)卫星快速移动导致用户在星载网络功能间频繁切换,频繁的用户上下文传输不仅会引发额外的能量消耗,还会增加信令流程完成时间;2)星载网络功能与用户间的距离动态变化,当星载网络功能远离用户时,信令和数据包的传输跳数增加,增加传输时间和转发开销;3)卫星覆盖广泛以及地球表面流量分布不均,导致星载网络功能负载剧烈波动,同时星载资源匮乏,使得卫星负载大时无法及时处理信令和转发数据。由于负责数据转发的用户面承载的建立、修改和删除过程依赖控制面信令流程,信令流程完成时间增加会导致网络响应变慢、连接不稳定等问题,降低用户服务质量。本文针对上述挑战,研究了在卫星网络中部署部分核心网功能,重点关注了星载核心网中的用户上下文管理、网络功能部署和信令处理等关键问题。本文的主要研究工作和创新性成果如下:1.星载核心网的用户上下文管理与部署策略研究为降低信令流程完成时间和信令转发开销,提出了星载网络功能间用户上下文传输机制和部署算法。首先,针对用户在星载网络功能间频繁切换,导致用户上下文传输时间增加的问题,提出了一种星载核心网架构。该架构通过解耦网络功能实例与状态信息,在低轨星座中部署无状态网络功能实例。接着,设计了星载网络功能之间的用户上下文传输机制。该机制中本地和远端数据库共同存储用户上下文,并采用异步方式进行同步,同时通过聚合批量用户上下文降低平均上下文获取时间。为了优化多种星载网络功能部署问题,综合考虑了部署成本、信令完成时间、以及转发开销,对星载网络功能部署问题进行建模。在此基础上,设计了一种星载网络功能部署算法,计算每种网络功能部署数量和位置。仿真结果表明,相较于传统单个用户上下文逐一传输的方法,基于无状态的批量上下文传输大幅度降低了网络功能实例获取用户上下文的平均时间。此外,与采用透明模式将信令转发至地面核心网相比,所提出的星载核心网在切换流程和服务请求流程中显著减少了信令流程完成时间和转发开销,同时保持了所有信令流程的高可靠性。2.星载用户面功能的服务区域分配与动态部署策略研究为克服星载用户面功能(User Plane Function,UPF)远离用户以及转发能力受限导致的用户面时延增加问题,提出了静态服务区域分配动态实例部署方案。首先,针对星载UPF负载剧烈波动,提出了一种静态服务区域分配算法,为每个星载UPF分配特定的地面区域,保证星载UPF负载的长期稳定,保障数据包转发速率和可靠性。接着,为应对星载UPF因远离静态服务区域而导致的数据包传输时间增加,利用可预测的星历信息和卫星运动规律,提出了星载UPF的迁移规则,并对星载UPF迁移过程建模。在此基础上,通过构建迁移图,设计了一种星载UPF迁移算法,使得星载UPF能够根据卫星网络的动态拓扑优化部署位置,降低用户面时延和迁移开销。为了解决卫星资源受限导致的星载UPF转发效率低下问题,基于矢量包处理技术设计并实现了高效的UPF原型。仿真结果表明,静态服务区域分配有效保障了星载UPF的负载稳定,动态部署使得星载UPF始终保持在服务区域附近,缓解了因星载UPF与用户间距离增加而导致的数据包传输时间增加。与采用负载均衡的静态实例部署动态服务区域分配方案相比,所提方案在平均时延上降低了 69.1%,能量消耗降低了 41.8%。3.星载控制面功能的协同信令处理与迁移策略研究为解决星载基站间切换流程完成时间增加导致的服务连续性下降问题,提出了协同信令处理方案和星载控制面功能迁移算法。首先,针对星载网络功能间的频繁信令交互,提出合并网络功能与处理逻辑,简化信令处理流程,减少信令开销。接着,提出了星地协同信令处理方案,将信令划分为时间敏感型信令和时间容忍型信令,结合信令的串行特性,设计星地用户上下文同步机制,实现星地协同信令处理。为了降低控制面功能迁移开销,设计了控制面功能迁移流程,基于批量用户的行为一致性与上下文相似性,采用数据压缩方法,减少迁移过程中的数据传输量。针对星载控制面功能远离用户导致的信令传输时间增加,基于深度强化学习提出了一种星载控制面功能迁移算法,综合考虑用户分布、星载UPF分布以及卫星网络状态,降低切换流程完成时间与迁移次数。仿真结果表明,与星上处理所有信令相比,星地协同信令处理节省了约67%的星上计算资源消耗,并在高负载时减少了 97.5%切换流程完成时间。设计的迁移流程仅传输了原始数据量的3%,而迁移算法通过减少迁移次数,实现了更高的迁移效率。更多还原

【Abstract】 With the continuous improvement of satellite performance,base station and key network functions of the core network are deployed on satellites,enabling the satellite network to take on partial core network functions.This enables the direct processing of signaling and data forwarding on the satellite,and avoiding the need to relay signaling and data to the ground.The onboard core network not only significantly reduces the transmission time of signaling and data packets but also effectively alleviates the signaling storm between the satellite and the gateway,enabling it to directly provide services to users without relying on terrestrial infrastructure.However,the infrastructure in terrestrial network is fixed,and their resources are more abundant compared to satellites.On the other hand,satellites are in high-speed motion and have limited resources,which makes it difficult for core network designed for terrestrial network to adapt to the dynamic nature of satellite network.The onboard core network faces the following three main challenges:1)The rapid movement of satellites leads to frequent handovers of users between onboard network functions.These frequent user context transmissions not only cause additional energy consumption but also increase signaling procedures completion time.2)The dynamic change in the distance between onboard network functions and users increases the number of transmission hops when the onboard network function moves farther from the user,leading to increased transmission time and forwarding overhead.3)The wide coverage of satellites and the uneven distribution of traffic on the Earth’s surface result in severe fluctuations in the load of onboard network functions.Moreover,the limited resources of satellites prevent timely signaling and data forwarding when the satellite is heavily loaded.Since the establishment,modification,and deletion of the user plane carrier responsible for data forwarding rely on control plane signaling procedures,an increase in the completion time of these signaling procedures leads to slower network responses,unstable connections,and other issues,thus degrading the quality of services.This thesis addresses these challenges by investigating the deployment of partial core network functions in satellite network,focusing on key issues such as user context management,network function deployment,and signaling processing in onboard core network.The main research work and key innovative contributions are as follows:1.Research on User Context Management and Deployment Strategy for the Onboard Core NetworkTo reduce signaling procedure completion time and signaling forwarding overhead,a user context transfer mechanism between onboard network functions and a deployment algorithm for network functions are proposed.First,to address the issue of increased user context transfer time caused by frequent handovers between onboard network functions,an onboard core network architecture is proposed.The architecture decouples network function instances from state information and deploys stateless network function instances in low Earth orbit(LEO)satellite constellations.Next,a user context transfer mechanism between onboard network functions is designed.In this mechanism,local and remote databases jointly store user contexts and synchronize asynchronously,while batch aggregation of user contexts is used to reduce the average context acquisition time.To optimize the deployment of multiple onboard network functions,the deployment problem is modeled by comprehensively considering the deployment cost,signaling completion time,and forwarding overhead.Based on this,an onboard network function deployment algorithm is designed to determine the optimal number and locations for deploying each network function.Experimental results show that,compared to the traditional method of transmitting user contexts individually,stateless batch context transfer significantly reduces the average time required for network function instances to acquire user contexts.Moreover,compared to the transparent mode of forwarding signaling to the terrestrial core network,the proposed onboard core network significantly reduces signaling procedure completion time and forwarding overhead in the handover and service request procedures,while maintaining high reliability for all signaling procedures.2.Research on Service Area Allocation and Dynamic Deployment Strategy for the Onboard User Plane FunctionTo overcome the problem of increased user plane latency caused by the onboard User Plane Function(UPF)being distant from users and its limited forwarding capacity,a scheme combining static service area allocation and dynamic instance deployment is proposed.First,to address the severe load fluctuations of the onboard UPF,a static service area allocation algorithm is proposed,which assigns specific terrestrial areas to each onboard UPF.This ensures long-term stability in the onboard UPF load and maintains packet forwarding rate and reliability.Next,to address the increased packet transmission time caused by the onboard UPF moving away from their static service areas,migration rulcs for the onboard UPF are proposed based on predictable ephemeris information and satellite motion patterns,and the migration process of the onboard UPF is modeled.Based on this,a migration graph is constructed,and a migration algorithm for the onboard UPF is designed,enabling the onboard UPF to optimize its deployment position according to the dynamic topology of the satellite network,thus reducing user plane latency and migration overhead.To address the low forwarding efficiency of the onboard UPF caused by limited satellite resources,an efficient UPF prototype based on vector packet processing technology is designed and implemented.Experimental results show that static service area allocation effeectively ensures the stability of the onboard UPF load,while dynamic deployment keeps the onboard UPF close to its service area,and alleviating the increased packet transmission time caused by greater distances between the onboard UPF and users.Compared to the scheme of static instance deployment with load balancing and dynamic service area allocation,the proposed scheme reduces average latency by 69.1%and energy consumption by 41.8%.3.Research on Collaborative Signaling Processing and Migration Strategy for the Onboard Control Plane FunctionTo address the issue of service continuity degradation caused by increased handover procedure completion time between onboard base stations,a collaborative signaling processing scheme and an onboard control plane function migration algorithm are proposed.First,to address frequent signaling interactions between onboard network functions,a method of merging network functions and processing logic is proposed,simplifying the signaling processing procedure and reducing signaling overhead.Next,a collaborative signaling processing scheme between satellite and ground is proposed,in which signaling is classified into time-sensitive and time-tolerant types.Leveraging the sequential characteristics of signaling procedures,a satellite-ground user context synchronization mechanism is designed to achieve collaborative signaling processing.To reduce control plane function migration overhead,a migration procedure is designed,employing data compression methods based on the behavior consistency and context similarity of batch users to reduce the amount of data transferred during migration.To address the increased signaling transmission time caused by the onboard control plane function being distant from users,a satellite control plane function migration algorithm based on deep reinforcement learning is proposed.It comprehensively considers user distribution,onboard UPF distribution,and satellite network status,thereby reducing handover procedure completion time and migration frequency.Experimental results show that,compared to processing all signaling onboard,satellite-ground collaborative signaling processing reduces onboard computational resource consumption by approximately 67%and decreases handover procedure completion time by 97.5%under high load.The designed migration procedure transmitted only 3%of the original data,while the migration algorithm improved efficiency by reducing the number of migrations.更多还原