一种基于时空联合感知的算力资源动态调度方法及系统
By employing a spatiotemporal joint perception-based dynamic scheduling method for computing resources, combined with graph neural networks, time series models, and reinforcement learning, the problem of spatiotemporal feature fusion and global optimization in computing resource scheduling under highly dynamic network environments is solved. This achieves efficient and balanced resource scheduling, improving system stability and task success rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TSINGHUA SHENZHEN INTERNATIONAL GRADUATE SCHOOL
- Filing Date
- 2026-05-13
- Publication Date
- 2026-07-17
AI Technical Summary
In highly dynamic network environments, existing computing resource scheduling methods struggle to achieve deep integration of spatiotemporal characteristics, and their optimization objectives lack global foresight, leading to uneven resource utilization and insufficient system stability, which fails to meet the low latency requirements of high-level autonomous driving and intelligent traffic management such as 5G-V2X.
A dynamic scheduling method for computing resources based on spatiotemporal joint perception is adopted. By nested fusion graph neural networks and time series models, combined with reinforcement learning agents, spatiotemporal feature extraction and scheduling decisions of computing networks are realized. A two-dimensional reward function is introduced to ensure the quality of real-time service of tasks and the long-term stability of the system.
It achieves efficient, balanced, and forward-looking scheduling of computing resources, improves task success rate and system resource utilization, ensures stability and reliability in complex network environments, and achieves a global and long-term optimal balance between service quality and resource efficiency.
Smart Images
Figure CN122179842B_ABST