Cost-Effective Mobility-Aware Multi-Connectivity Scheme for Physical-to-Virtual Communications in Digital Twin Networks

Maintaining robust physical-to-virtual twin connectivity remains a critical challenge in the deployment of highfidelity digital twin networks (DTNs). To ensure that each virtual twin (VT) in such DTNs is a true replica of its counterpart physical twin (PT), a reliable and cost-effective connectivity framework is essential, especially considering the potential mobility of PTs in the physical environment. This paper thus presents a new multi-modal PT trajectory prediction-enabled multi-connectivity framework where the counterpart VT model is proactively duplicated along multiple predicted paths to compensate for the effects of prediction errors, thereby avoiding communication loss between such PT-VT pair. While such a proposed scheme can improve the overall reliability of the network, the resulting connectivity solution may suffer from network congestion issues, complicating the overall cost. To address this, we analyze the trade-offs between overall reliability and cost and formulate an optimal VT placement problem aimed at minimizing delay, longterm connectivity cost, and prediction errors. We then propose a deep reinforcement learning-based adaptive VT placement strategy for cost-effective, mobility-aware PT-VT connectivity. Simulation results show the solution effectively meets the reliability and cost requirements of DTNs.