Two-Level ISTN Intelligent Resource Management: Macro Multipath Planning and Micro A3C-Based Scheduling

Integrated satellite-terrestrial networks (ISTNs) are pivotal for ubiquitous 5G/6G connectivity but face severe resource allocation and traffic scheduling challenges due to highly dynamic satellite orbits and heterogeneous traffic demands. To address these issues, this paper proposes a software-defined networking (SDN)-based ISTN architecture featuring a novel two-level temporal hierarchy that effectively decouples long-term multi-path planning from real-time traffic scheduling. First, regarding long-term planning, we formulate a time-aware particle swarm optimization (PSO) algorithm incorporating a penalty mechanism to enforce link-duration constraints derived from orbital dynamics. Second, for real-time scheduling, we design a distributed asynchronous advantage actor-critic (A3C) framework. This hybrid mechanism combines local reinforcement learning at user proxies with lightweight global supervision, thereby mitigating centralized computational bottlenecks and ensuring scalable coordination. Simulation results demonstrate that the proposed method significantly outperforms state-of-the-art techniques in terms of response time, load balancing, and packet delivery reliability.