Ns3 meets Sionna: Using Realistic Channels in Network Simulation

TL;DR

Ns3Sionna addresses the gap between fast network simulators and physically accurate wireless channels by integrating a ray-tracing-based propagation model (Sionna RT) into ns-3. The approach leverages environment-specific 3D scenes, a 3D mobility model, and a caching strategy based on coherence time $T_C$ to avoid recomputing channels, while offloading P2MP calculations to Sionna and GPUs for parallelism. It demonstrates indoor and outdoor scenarios with improved path loss, propagation delay, and CFR realism compared with traditional ns-3 models, including spatial and temporal correlation and fine-grained channel state information. The results show substantial CPU/GPU speedups through predictive calculations and caching, with limitations in very large/dense mobile scenes and clear pathways to extend to MIMO and RIS in future work.

Abstract

Network simulators are indispensable tools for the advancement of wireless network technologies, offering a cost-effective and controlled environment to simulate real-world network behavior. However, traditional simulators, such as the widely used ns-3, exhibit limitations in accurately modeling indoor and outdoor scenarios due to their reliance on simplified statistical and stochastic channel propagation models, which often fail to accurately capture physical phenomena like multipath signal propagation and shadowing by obstacles in the line-of-sight path. We present Ns3Sionna, which integrates a ray tracing-based channel model, implemented using the Sionna RT framework, within the ns-3 network simulator. It allows to simulate environment-specific and physically accurate channel realizations for a given 3D scene and wireless device positions. Additionally, a mobility model based on ray tracing was developed to accurately represent device movements within the simulated 3D space. Ns3Sionna provides more realistic path and delay loss estimates for both indoor and outdoor environments than existing ns-3 propagation models, particularly in terms of spatial and temporal correlation. Moreover, fine-grained channel state information is provided, which could be used for the development of sensing applications. Due to the significant computational demands of ray tracing, Ns3Sionna takes advantage of the parallel execution capabilities of modern GPUs and multi-core CPUs by incorporating intelligent pre-caching mechanisms that leverage the channel's coherence time to optimize runtime performance. This enables the efficient simulation of scenarios with a small to medium number of mobile nodes.

Figures (11)

• Figure 1: Channel computation time in Sionna RT
• Figure 2: To speed up execution Ns3Sionna converts future node locations into locations of virtual nodes which can be calculated in parallel.
• Figure 3: Architecture of the Ns3Sionna
• Figure 4: Validation of path loss and propagation delay in free-space
• Figure 5: Indoor scenario - two rooms and open door