Outdoor Bistatic 5G ISAC Dataset from POWDER: Raw Uplink SRS with Vehicle GNSS and Drone-Video Ground Truth

What this is Raw 5G uplink sounding reference signals (SRS) received by a base station, recorded for sensing. In two runs, a 5G NR base station received the SRS of two parked UEs (Quectel RM520N-GL cellular modems on car roofs) every 2.5 ms, on a four-port panel overlooking a parking lot. Every SRS was recorded raw: the received resource elements on all four receive ports, the reference sequences, and the exact air time. Meanwhile a car carrying a GNSS receiver drove a planned route through the lot, and a drone filmed the lot from a hover and then flew low passes. Each SRS travels from a UE to the antenna panel along the direct path and, more faintly, by way of everything that reflects it. A moving reflector (the car, the drone, people, other cars) adds an echo whose delay and Doppler shift change over time. That makes the record a bistatic sensing dataset, with two transmitters and one four-port receiver. The ground truth is the car's GNSS, the drone's flight log, and the video with a timed camera pose for every frame. The data were collected on POWDER, the Platform for Open Wireless Data-driven Experimental Research (https://powderwireless.net): a city-scale, remotely accessible wireless testbed at the University of Utah in Salt Lake City, one of the NSF's Platforms for Advanced Wireless Research (PAWR). The base station is POWDER's: the radio unit and receive panel on the Merrill Engineering Building's roof, their GNSS-locked timing, and the server that ran the gNB. Why: to let researchers develop and test integrated sensing and communication (ISAC) methods on data from a real 5G base station: detecting, locating and tracking moving objects from the uplink sounding reference signals that phones send anyway. Raw SRS from a standard gNB, with independent ground truth, is rarely public. What is not in it: detections, tracks, positions estimated from the radio, aligned or cleaned channels, labels, or any other processing result. Those are for users of the data to produce, so that different methods can be compared on the same raw signals. At a glance Platform POWDER, the University of Utah's city-scale wireless testbed (NSF PAWR): its rooftop radio unit and antenna panel at the Merrill Engineering Building, and a POWDER server running the gNB Place Merrill Engineering Lot, University of Utah, Salt Lake City (40.7703 N, 111.8455 W), an open lot of ~180 × 140 m Date 2026-10-25, 23:23 to 00:06 UTC (late afternoon local time, low sun) Runs two (run1, run2), 13.1 and 15.6 minutes of SRS Base station ocudu gNB on a PTP-disciplined host; Benetel RAN650 radio unit; band n78, 3.42 GHz carrier, 30 kHz subcarriers, TDD DDDSU (2.5 ms) Receive antenna L-com HG3517DP4-090 dual-slant panel, 16.5 dBi, 95° × 6.5°, two columns × ±45°; four ports = Rx 0–3; 16.3 m above the ground at the mast (1466.3 m NAVD88), on the Merrill Engineering Building's roof, south of the lot Transmitters two UEs (Quectel RM520N-GL) parked in the lot, antennas on car roofs at 1.75 m: UE-A (Taoglas TGX.04) and UE-B (Taoglas MA963) SRS every 5 slots (2.5 ms) per UE; 2 Tx ports (cyclic shifts 0 and 6 of 12) on one comb-4 RE set; 288 REs per port at 120 kHz spacing (34.56 MHz); full power (p0 −50 dBm) Snapshots 1.37 million: run 1 313,379 (UE-A) and 313,380 (UE-B); run 2 369,538 (UE-A) and 373,574 (UE-B); none lost Delay resolution one tap = 1 / 34.56 MHz = 28.9 ns, 8.7 m of path length Target car a black 2022 Honda CR-V Hybrid (4.63 × 1.85 × 1.69 m); u-blox ZED-X20P GNSS, 10 Hz, standalone (no RTK); a planned route through the aisles at up to 6 m/s (14 mph); GNSS logs of 7.6 and 8.3 minutes, each ending with the car parked at site (40, 53) m, 4.7 and 6.3 minutes before the last SRS Drone DJI Air 2S (1-inch 20 MP sensor; 22 mm equivalent f/2.8 lens, 88° diagonal): a 3840 × 2160, 29.97 fps H.265 video of each run (hover at 50 m / 62 m above takeoff, then a descent with low passes and the landing); a ~10 Hz flight log Licence CC BY 4.0 (LICENSE.md); the OpenStreetMap extract stays under the ODbL Uses Possible uses: bistatic detection, delay and Doppler estimation, localisation and tracking from one or two transmitters, angle of arrival with a two-column panel, static-scene and clutter modelling, channel modelling at 3.4 GHz, the timing of a 5G receiver against GNSS, and radio-vision fusion with the timed video. Already used for: the authors' check that the record is usable: moving-target detection on the delay-Doppler map, scored against the car's GNSS and the drone's log; the receive-port angle; the clock agreement. None of these results are in the record. Splits: none given. Run 1 and run 2 follow the same route and can serve as development and test sets. Limits: one site, one evening, two short runs, one target car and one drone; standalone GNSS truth (decimetre to metre accuracy); a panel orientation known to a few degrees. Results should not be generalised to other sites without care. Not for: identifying or tracking individual people or vehicles. Authors, citation, funding and contact Authors: Dustin Maas, Xiangyao Tang, Mahnoor Fatima Saad, Muhammad Basit Iqbal Awan and Jacobus Van der Merwe, all at the University of Utah. The data were recorded on POWDER. Cite as: Dustin Maas, Xiangyao Tang, Mahnoor Fatima Saad, Muhammad Basit Iqbal Awan and Jacobus Van der Merwe (2026). Outdoor Bistatic 5G ISAC Dataset from POWDER: Raw Uplink SRS with Vehicle GNSS and Drone-Video Ground Truth [data set]. Zenodo. https://doi.org/10.5281/zenodo.23717707 Please also cite POWDER: J. Breen et al., "POWDER: Platform for Open Wireless Data-driven Experimental Research," Computer Networks, vol. 197, 108281, 2021, https://doi.org/10.1016/j.comnet.2021.108281. Funding: This material is based upon work supported by the National Science Foundation under Grant No. 2346555 (CIRC: ENS/Grand: POWDER-ENS - Enhancing and Sustaining the POWDER Platform). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. Contact: Dustin Maas, [e-mail].