Toward Digital Network Twins: Integrating Sionna RT in ns-3 for 6G Multi-RAT Networks Simulations

TL;DR

The paper tackles the need for high-fidelity, open-source tools for 6G network research by integrating Sionna RT, a deterministic ray-tracing engine, with ns-3 to create a full-stack, multi-RAT Digital Network Twin. The proposed disaggregated architecture allows the ray tracer to run on external GPUs or cloud servers, communicating via UDP with ns-3, and employs caching and grouped calculations to maintain performance. Through a vehicular urban scenario, the study demonstrates substantial improvements in channel realism over traditional stochastic models, reporting up to around $65\%$ differences in application-layer results and up to $56\%$ discrepancies at higher frequency bands, highlighting ray tracing's critical role in realistic 6G simulations. The work outlines clear research directions toward scalable architectures, hardware acceleration, and richer RT-derived data usage to advance training, optimization, and management of future networks.

Abstract

The increasing complexity of 6G systems demands innovative tools for network management, simulation, and optimization. This work introduces the integration of ns-3 with Sionna RT, establishing the foundation for the first open source full-stack Digital Network Twin (DNT) capable of supporting multi-RAT. By incorporating a deterministic ray tracer for precise and site-specific channel modeling, this framework addresses limitations of traditional stochastic models and enables realistic, dynamic, and multilayered wireless network simulations. Tested in a challenging vehicular urban scenario, the proposed solution demonstrates significant improvements in accurately modeling wireless channels and their cascading effects on higher network layers. With up to 65% observed differences in application-layer performance compared to stochastic models, this work highlights the transformative potential of ray-traced simulations for 6G research, training, and network management.

Figures (7)

• Figure 1: Setup overview with procedures pseudo-code.
• Figure 2: The considered 3D vehicular scenario, where moving vehicles are randomly connected or not. For testing purposes, every vehicle was supposed to be .
• Figure 3: Computed path gain between two entities using three at the same carrier at 5.89 GHz. The highlighted sections of the graphs represent measurements.
• Figure 4: Impact of ray tracing on successfully received across different at 5.89 GHz.
• Figure 5: Packet Reception Disagreement Ratio (PRDR), percentage of messages marked as delivered by ns-3 stochastic models incoherently with the ray tracer channel.