Integrate Sionna RT Ray Tracing for Physics-Based Coverage Simulation #2
Description
Integrate Sionna RT Ray Tracing for Physics-Based Coverage Simulation
Is your feature request related to a problem? Please describe.
SmartRAN Studio currently uses Sionna's 3GPP TR 38.901 UMa statistical channel model for RF propagation and coverage simulation. While this is fast and works well for large-scale scenarios, it's not as physically accurate for real-world RF data generation. Statistical models don't account for actual buildings, terrain, reflections, diffraction, and other real-world propagation effects. When I'm simulating user coverage, I need it to be based on ray tracing that models how radio waves actually propagate through real environments, not statistical approximations.
Describe the solution you'd like
Integrate Sionna RT (ray tracing module) as the propagation engine for coverage simulation. This would allow:
- Loading 3D scenes with actual building geometry and terrain
- Computing coverage based on physics-based ray tracing instead of statistical models
- Getting realistic propagation that accounts for reflections, diffraction, and obstacles
- Option to use ray tracing mode when accuracy matters, keeping the fast statistical mode for quick iterations
Basically, when I run srs sim compute, I want to be able to specify that it should use ray tracing with a 3D scene instead of the statistical UMa model.
Describe alternatives you've considered
- Keep using the statistical model, but it's as not accurate enough for RF simulation data
- Statistical model will act as all individual antenna elements together put together, but I want to include individual rays for more granular analysis.
Additional context
Platform Compatibility:
From what I've researched, Sionna RT depends on NVIDIA OptiX for ray tracing acceleration. This has platform implications:
- Native Linux: OptiX is fully supported and should work properly
- WSL2: OptiX is NOT officially supported on WSL2 - users report errors like
OPTIX_ERROR_LIBRARY_NOT_FOUND. There are unofficial workarounds but they're unreliable.
Implementation Priority:
Since many SmartRAN Studio users may be on WSL2, I think we should:
- First target native Linux - get it working properly on the fully supported platform
- Document the WSL2 limitations clearly - let users know they need native Linux or cloud instances for ray tracing features
- Provide deployment alternatives - document how WSL2 users can set up dual-boot Linux or use cloud GPU instances
What needs to be figured out:
- What exact versions of Sionna, OptiX, CUDA, drivers, etc. are needed?
- How to package this in the Docker container (especially since OptiX won't work in WSL2 anyway)?
- How to load and manage 3D scenes?
- What GPU hardware is required (RT cores needed? minimum VRAM?)?
- How to integrate this into the existing
MultiCellSim.compute()workflow?
Documentation needed:
- Installation guide for native Linux (step-by-step OptiX setup)
- Clear platform compatibility statement (native Linux required, WSL2 not supported)
- Alternative deployment options for Windows users
- Tutorial on creating/loading 3D scenes
- Example scenes to get started
Resources:
- Sionna RT docs: https://nvlabs.github.io/sionna/rt/index.html
- OptiX downloads: https://developer.nvidia.com/designworks/optix/downloads
- WSL2 OptiX issues: https://forums.developer.nvidia.com/t/problem-running-optix-7-6-in-wsl/239355
The main goal is to get physically accurate, ray-traced coverage simulation working in SmartRAN Studio, with Linux as the primary supported platform.