Speaker
Description
We present gr-rt_channel_emulator, a GNU Radio out-of-tree module for position-aware, ray-tracing-assisted channel emulation with runtime updates. The module pairs GNU Radio's real-time IQ processing with NVIDIA Sionna RT as a decoupled channel-impulse-response (CIR) engine. A 3D scene in Mitsuba XML is loaded, transmitter and receiver nodes are placed at configurable coordinates, and Sionna RT computes propagation paths for a chosen carrier frequency, antenna configuration, polarization, and interaction depth. The resulting paths are reduced to a discrete complex baseband CIR and applied to the streaming IQ inside a hierarchical block built around GNU Radio's native channel model, so standard impairments — additive noise, carrier-frequency offset, timing offset, and scaling — remain available alongside the ray-traced response. Because ray tracing is far slower than the sample rate, CIR computation runs off the streaming path and updated taps are swapped in asynchronously, keeping the flowgraph real-time.
A central feature is dynamic node positioning. Transmitter and receiver coordinates update at runtime — fed to the ray tracer and delivered to the flowgraph as CIRs through GNU Radio message ports — letting the emulated channel track motion without rebuilding the flowgraph. Position streams can originate from external autonomy and robotics frameworks such as ROS or QGroundControl, enabling experiments in which mobile nodes, robots, or unmanned aerial systems move through a 3D environment while the channel is recomputed and applied to the live IQ stream.
Unlike ray-tracing-driven emulators targeting network simulators (ns-3) or cellular stacks (OpenAirInterface), and unlike testbed emulators that rely on simplified geometric models, gr-rt_channel_emulator brings full ray-traced responses natively into GNU Radio flowgraphs with autonomy-driven mobility. Each IQ stream carries its geometry, materials, carrier frequency, antenna settings, node locations, and ground-truth ray-traced taps, yielding labeled datasets for channel estimation, signal classification, RF fingerprinting, physical-layer security, and other AI-assisted PHY tasks. The module ships with ready-to-use FAU, AERPAW, and POWDER scenes and supports custom indoor and outdoor environments.
| Talk Length | 15 Minutes |
|---|---|
| Acknowledge | Acknowledge In-Person |