Hierarchical Digital Twin for Efficient 6G Network Orchestration via Adaptive Attribute Selection and Scalable Network Modeling

Achieving both a holistic and in-depth understanding of network dynamics through accurate modeling is essential for orchestrating future 6G networks, considering their increasing complexity and service diversity. However, traditional situation-agnostic data collection and network modeling approaches often undermine the efficacy and timeliness of network orchestration in such complex environments. Furthermore, temporal misalignments caused by varying modeling delays across distributed networks further impair centralized decision-making. To address these challenges, this paper proposes a hierarchical digital twin framework with an adaptive layered architecture designed for problem-oriented 6G network modeling and orchestration. At higher layers, we introduce an adaptive attribute selection mechanism that efficiently evaluates network situations and identifies problematic areas. This mechanism prioritizes critical attributes by jointly considering their relevance to current network objectives and modeling complexity. At lower layers, these prioritized attributes and critical users are selectively incorporated into scalable network modeling. More detailed digital twins are then created to deliver targeted solutions for optimizing user association and power allocation. Additionally, we implement a multi-level synchronization mechanism to ensure temporal alignment among the digital twins, thereby enhancing the effectiveness of model-based orchestration. Extensive simulations validate the efficient identification of pressing operational issues and the effective orchestration of complex 6G networks.