MAGNETAR: Joint RF Pose Inference
I lead MAGNETAR from probabilistic formulation and neural architecture design through real-to-sim calibration, MobiFR3 integration, experiment design, and evaluation. Given a known room layout and receiver pose, the method estimates a joint posterior over a transmitter’s planar position and heading from one asynchronous RF multipath snapshot, retaining multiple physically plausible poses.
The three-minute overview shows why position and heading must be inferred together, how MAGNETAR scores pose hypotheses, and how it is evaluated with robotic measurements.
Watch the 45-second Brooklyn 6G Summit 2026 research showcase.


Method
- Represents each observation by angle-of-arrival and signal-to-noise-ratio estimates, together with the known room layout and receiver pose.
- Uses a heading-conditioned shared 2D U-Net to score and jointly normalize a discretized position-heading grid.
- Trains primarily in real-to-sim-calibrated 10-GHz simulation, with measured antenna patterns and RF-chain/noise effects, then augments training with 1,800 measured records.
- Fuses joint posteriors across observations to accumulate compatible position-and-heading evidence from multiple receiver viewpoints.
Physical evaluation
The experimental study uses MobiFR3, the robotic RF testbed that I develop with collaborators. On 24,000 held-out measurements, the heading-conditioned scorer achieved the lowest real-data negative log-likelihood and the highest joint hit rate among the tested models. I designed the inference model, calibration, experimental protocol, and evaluation, and coordinated RF–robot integration with student collaborators working on RF operation and data acquisition.
Paper: MAGNETAR: Multipath-Guided Spatial Posteriors for Transmitter Pose Inference in the Upper Mid-Band — first and corresponding author; IEEE ICRA 2027, under review.