Joint VNF Deployment and Information Synchronization in Digital Twin Driven Network Slicing via Deep Reinforcement Learning

Network slicing (NS) provides customized services for Internet of Vehicles (IoV) users by creating logical virtual networks. Digital Twin Network (DTN) enables IoV network monitoring and low-cost policy validation. However, it needs to consider reducing the delay in transmitting synchronization information from the physical layer to the digital twin layer. Therefore, this paper proposes a method based on Digital Twin Driven Network Slicing (DTDNS) for Virtual Network Function (VNF) deployment and network information synchronization. Firstly, we abstract the data collection function of DTDNS as an information synchronization VNF, deploy it jointly with service VNFs, and propose a VNFs joint deployment model and an information synchronization model. Secondly, we introduce an optimization problem to maximize service and information synchronization utility, which consists of a deployment subproblem and an information synchronization subproblem. Additionally, we introduce a distributed VNF deployment and information synchronization algorithm to address these issues. Simulation results demonstrate that our proposed algorithm can reduce information synchronization delay and node deployment costs. Furthermore, the distributed VNF deployment and information synchronization algorithm can enhance algorithm convergence speed and reduce training time.