A Dual Time-Scale Offloading Strategy for Vehicle Digital Twins via Cloud-Edge-End Collaboration

With the advancement of vehicular digital twin (DT) technologies, managing data transmission and task offloading among vehicles, edge servers, and the cloud remains challenging. This paper presents a Double Time-Scale Hierarchical Soft Actor-Critic (DTH-SAC) strategy for collaborative digital twin offloading, addressing the trade-offs among model accuracy, latency, and energy consumption in vehicular networks. The long time scale optimizes generalized model deployment and edge server selection for global stability, while the short time scale adjusts data upload frequency, offloading decisions, and resource allocation for real-time updates. By facilitating edge server collaboration, DTH-SAC effectively mitigates issues from frequent base station handovers, maintaining model accuracy and consistency. Experimental results show that DTH-SAC outperforms SAC and PSO in accuracy, energy efficiency, and latency, providing a flexible solution for intelligent computing in vehicular networks.