A Digital Twin-Enabled Approach for Resource Allocation in Cross-Domain Network Slicing

Network Digital Twin (NDT) and network slicing are key technologies for the next-generation mobile networks (6G), enabling a transformation in network planning, construction, and operation. However, existing 5G networks face issues such as heterogeneous resource types, differing management strategies among operators, and cross-domain data privacy constraints, which hinder the interoperability of node resources and state information. This paper proposes the use of a DT network to create virtual counterparts of physical nodes, enabling cross-domain information exchange and collaborative optimization across Radio Access Network (RAN) and Core Network (CN) domains. The main contributions are: (1) a DT-based cross-domain heterogeneous network slicing architecture, using Software Defined Network (SDN) to mirror the operating states of physical nodes for interaction and collaboration between slices and twins; (2) a physical network model for pretraining resource scheduling strategies at the physical network layer; (3) a transfer learning mechanism to migrate pretrained strategy parameters from the physical to the virtual twin network, establishing collaborative resource allocation between the twin and physical networks. This mechanism further supports coordination between AI-RAN and the CN. Simulation results show that the proposed solution outperforms the baseline in multi-agent resource scheduling tasks, demonstrating significant improvements in cross-domain collaboration and joint resource scheduling in a multi-slice heterogeneous environment.