COLLABORATIVE LOCAL-GLOBAL MULTI-OBJECTIVE OPTIMIZATION FOR DIGITAL-TWIN COMPUTING POWER NETWORKS

The explosive growth of end devices and mobile applications has highlighted the challenges in task offloading and resource scheduling in computing power networks. Applying digital twin to computing power networks enables the use of real-time simulation environments to optimize task offloading and resource scheduling. Most decision-making approaches focus on centralized scheduling frameworks, which often suffer from scalability challenges. As the complexity of offloading scheduling problems increases, the search performance of these approaches may decrease. Therefore, we propose a hierarchical scheduling framework for digital-twin computing power networks that enhances the scalability of the system through the collaboration of local and global agents. Moreover, we design a two-stage multi-objective optimization algorithm in the hierarchical scheduling framework to jointly optimize task offloading and resource scheduling. Compared to three baseline scheduling algorithms, the proposed method achieves a 10.2% performance improvement.