Deployment and Synchronization for Reputation-Driven Vehicular Digital Twins: A Learning-Based Approach

With the rapid advancement of the Internet of Vehicles (IoV), the integration of Digital Twin (DT) technology has shown considerable potential across a wide range of application scenarios. However, in high-density traffic environments, the overlapping coverage of edge nodes presents a challenge for vehicles in selecting appropriate nodes for DT deployment. Additionally, heterogeneous edge node reputations may result in suboptimal deployments, leading to issues such as resource imbalance and synchronization failures. To address these challenges, we propose a DT-assisted vehicular edge computing framework. By evaluating contributions of Roadside Units (RSUs) during the synchronization, we develop a reputation-driven DT deployment model. In this model, RSUs with higher reputation can receive greater deployment rewards, thereby incentivizing their active participation in the DT deployment. Furthermore, to minimize the system Age of Information and deployment cost, we develop a Transformer-Critic-based Multi-Agent Proximal Policy Optimization (TC-MAPPO) algorithm to jointly optimize DT placement and resource allocation. Extensive experiments demonstrate the effectiveness of the proposed TC-MAPPO approach and highlight its superiority in enhancing overall average reward.