Repository navigation
Releases
Releases · le-liang/Multimodal-Wireless
v1.1.1-Sub-6G Data and Script Updated
In this release, we updated the sub-6G data-dumping script, along with the sub-6G v2i and v2v data to the public dataset.
For sub-6G with sufficient scattering and diffraction, we choose to activate the scattering and diffraction in Sionna. We choose the following parameters:
- We use
max_depth=1because direct and single-interaction paths accounted for approximately 99.25% of the power within the strongest 20 paths in our pilot scene. - Diffraction and diffuse scattering are enabled to capture propagation beyond specular reflections.
- To limit storage, we retain at most the 20 strongest paths per link and configuration, accepting some loss of delay-spread information.
- We use
scattering_coefficient=0.3as an experimental assumption, not a calibrated or ITU-recommended material property.
v1.1.0 — 🚀 20x Accelerated Data Generation
🚀 v1.1.0 — Accelerated Data Generation
- Upgraded pipeline: Reuse the static scene and update dynamic vehicle poses frame by frame in Sionna RT, replacing the previous workflow of exporting and loading a complete scene for every frame.
- 20× faster generation: The pipeline redesign, together with code-level optimizations, reduces total generation time by approximately 95% for the same workload while preserving the generated results.
- Lower storage requirements: Eliminate the need to generate and store a separate full scene for every frame, removing the large intermediate scene-storage overhead of long sequences.
- Bug fix: Address a minor issue reported by issue.
Note: Any differences in the generated channels between the original version and v1.1.0 arise solely from this issue fix. The pipeline architecture update itself does not change the channel results.
See the detailed pipeline comparison for the original and accelerated workflows.
v1.0.0 — Original Generation Pipeline
Release v1.0.0 stores the original version of the codes for Multimodal-Wireless dataset.