Joint Optimization of Vehicular Sensing and Vehicle Digital Twins Deployment for DT-Assisted IoVs

Emerging Intelligent Transportation Systems (ITS) applications in the B5G/6G era have higher demands on real-time information in real traffic scenarios, and the current Internet of Vehicles (IoVs) can hardly support the operation of such applications. To realize the efficient operation of ITS applications, we integrate Digital Twin (DT) technology into IoVs and propose a framework of DT-assisted cloud-edge collaboration IoVs for intelligent transportation. DT synchronization requires vehicular sensing and data uploading, where the sensing capability and sensing policy of different vehicles affect the accuracy of DT, and the simultaneous sensing of neighboring vehicles creates data redundancy. The deployment strategy of Vehicle DTs (VDTs) at the network edge affects the real-time performance of DT. The mobility of vehicles, the low-latency requirements of DT, and the limited heterogeneous resources of edge servers pose great challenges to the deployment of VDTs. To address the above problems, we established the vehicular sensing model considering sensing quality, cost and redundancy. Then, A DT synchronization mechanism is designed and an improved Age of Information (AoI) metric is used to measure the freshness of the real vehicle state data received by the cloud during the DT synchronization process. We proposed a joint optimization problem of vehicular sensing and VDTs deployment to maximize the system Quality of Services (QoS), which is reflected through the vehicular sensing quality, AoI and system cost in DT synchronization process. We develop an algorithm based on Multi-Agent Deep Reinforcement Learning (MADRL) to solve this optimization problem, called DTSD-MAPPO. Numerical results show that the scheme reduces the system cost and improves the accuracy and real-time in DT synchronization.