端到端性能保障的空天地融合网络非独占资源部署

空天地融合网络具有传输链路动态时变、节点性能差异大、网络结构复杂等特点，保障端到端性能的传输及计算资源合理分配面临挑战。为实现网络资源与业务需求的合理匹配，本项目以网络资源高效利用为目标，在允许资源竞争的条件下研究空天地融合网络中保障端到端性能的资源分配新方法。基于集中加分布的两层式网络资源管控架构，本项目拟利用网络积分方法，构建资源非独占条件下的网络资源分配与端到端传输性能之间的映射关系，并在此基础上采用网络化多智能体约束强化学习方法，在严格保证服务质量的基础上，实现高效的网络资源分配。通过项目研究，将面向空天地融合网络提出满足多样化端到端性能需求的异构无线接入及网络资源调度方法，为空天地融合网络中的端到端性能保障提供理论指导。.

The air-space-ground integrated networks are characterized by dynamic and time-varying transmission links, large node performance differences, and complex network structure, which make it challenging to ensure the reasonable allocation of transmission and processing resources for guaranteed end-to-end performance. To achieve a reasonable match between network resources and service demands, this project aims at efficient utilization of network resources and proposes a new method of resource allocation to guarantee end-to-end performance in air-space-ground integrated networks while allowing resource competition. Based on the centralized-plus-distributed two-tier network resource management architecture, this project proposes to use the network calculus method to construct a mapping relationship between network resource allocation and end-to-end transmission performance under the non-exclusive resource allocations, and on this basis, adopt a networked multi-agent constrained reinforcement learning method to achieve efficient network resource allocation with strictly guaranteed service quality. The project research will propose a heterogeneous wireless access and network resource scheduling method to meet the end-to-end performance of diverse requirements, and provide theoretical guidance for the end-to-end performance guarantee in air-space-ground integrated networks.