Experiments in Wireless Digital Twin Creation

The constant evolution and need for rapid deployment of more efficient and reliable wireless networks, including 5G and 6G, demands resource-heavy solutions to model and compute. Wireless digital twin systems (WDTS) have emerged as a critical tool to virtually replicate the physical conditions in a virtual environment, allowing for network optimization in real-time. However, traditional WDTS face significant limitations with reliance on resource-intensive software, exhaustive measurements, rigid training models, and detailed environment data. This paper investigates various methods to optimize WDTS in terms of efficiency and accuracy through exploring different software solutions and modeling techniques. Building on an evaluation of traditional approaches, a quantitative analysis of conventional wireless digital twin (WDT) methods is conducted, specifically with ray-tracing in WirelessInSite and NVIDIA Sionna, highlighting their strengths and limitations in complex and dynamic environments. Additionally, the effects of geometry modifications using the ScanNet++ dataset and Blender 3D modeling software is performed to propose a method of geometry simplification, improving simulation speed without sacrificing accuracy of results.