Digital Twin-Assisted Site-Specific Radio Resource Management in Vehicular Networks

This paper presents a practical high-fidelity digital twin (DT) for site-specific radio resource management in vehicular networks. Using NVIDIA Sionna™, an open-source library for link-level simulations, and its differentiable ray tracing tool, the proposed DT integrates detailed radio-frequency (RF) wave propagation, 3D environment and vehicles, and realistic vehicle mobility. The proposed DT enables the acquisition of global channel state information (CSI) in large-scale vehicular networks, facilitating accurate coverage map generation, signal-to-interference-plus-noise ratio (SINR) calculation, beamforming, and near real-time radio resource management optimization. To showcase the DT’s effectiveness in site-specific interference management, an integer programming problem is formulated to jointly optimize roadside unit (RSU) beamforming and vehicle-RSU association, maximizing the network’s sum rate. For performance evaluation, a high-fidelity DT of a realistic micro-urban scenario is considered, and the evaluated parameters were the path gain, sum rate, number of vehicles, maximum number of vehicles per RSU, and time. The advantages of the proposed DT over the stochastic 3GPP channel model are shown through coverage maps and path gain simulations in different scenarios. Simulation results also confirm that the proposed DT-assisted optimization approach provides up to 380% higher sum rate than the baseline approach. It is also shown that the proposed DT-assisted optimization requires a small latency of a few seconds for decision-making, providing scalable solutions for large-scale vehicular networks.