Robust indoor positioning using 5G NR Positioning Reference Signals (PRS) with multipath rejection and Kalman filtering.
12 gNodeBs in a ray-traced train station with UE trajectory
| Metric | RANSAC-only | RANSAC + UKF |
|---|---|---|
| Median error | 0.67 m | 1.23 m |
| 95th percentile | 2.41 m | 1.85 m |
23% reduction in worst-case error — UKF trades slight median bias for tighter tails.
PRS Signals → TDoA Estimation → RANSAC Outlier Rejection → UKF Tracking → Position
- TDoA Estimation — Correlate 5G PRS with local replicas
- RANSAC — Sample anchor triplets, reject NLOS outliers
- Gauss-Newton — Solve hyperbolic intersection on inliers
- UKF — Constant-velocity prior smooths trajectory
| Parameter | Value |
|---|---|
| Transmitters | 12 gNodeBs |
| Carrier | 3.5 GHz |
| Bandwidth | 100 RB PRS (30 kHz SCS) |
| Channel | Ray tracing (2 refl, 1 diffr) |
| SNR range | -10 to 25 dB |
| Trials | 1,300 Monte Carlo |
% Configure in PositioningConfig.m, then run:
main_5g_positioning# Generate plots
python generate_positioning_plots.pyMATLAB 5G Toolbox Ray Tracing RANSAC UKF Python
Research project @ TU Hamburg, Institut für Hochfrequenztechnik