Session F6a: Open-Source Data and Tools for PNT Research and Development

BUC-NAV: Urban GNSS/INS Dataset with Tactical-Grade IMU and Multi-Receiver Architecture
Radu-Andrei Cioaca, University Politehnica of Bucharest; Florin Stoican, Three Tensors S.R.L. and University Politehnica of Bucharest; Alexandru Pandele, Andrei-Alexandru Marinache, Dragos Munteanu, and Mihail-Remus Palade, Romanian InSpace Engineering S.R.L.; Cristian Rusu and Paul Irofti, Three Tensors S.R.L. and University of Bucharest; Levente Tamas, Technical University of Cluj-Napoca; Gianluca Caparra, European Space Agency (ESA), ESTEC
Location: Windsong 7-8
Alternate Number 1

Reliable positioning in dense urban environments remains difficult for GNSS due to signal blockage, non-line-of-sight reception, and persistent multipath, motivating fusion with inertial sensing and the development of advanced estimators that operate on raw observables. Progress on such methods depends on datasets that contain synchronized raw GNSS measurements together with a high-grade inertial reference, yet most public datasets rely on automotive- or consumer-grade IMUs, cover short durations, and concentrate on a few well-studied locations. We present BUC-NAV, an urban GNSS/INS dataset collected in Bucharest, Romania, a dense Eastern European area under-represented in the navigation literature. Its core collection is built around four features: a tactical-grade IMU; a shared-antenna, multi-receiver front end that places a professional Septentrio mosaic-X5 and a dual-frequency u-blox F9P under identical radio-frequency conditions for fair comparison; more than 25 hours of recording across 16 sessions spanning suburban to deep-urban conditions, including trajectories exceeding two hours; and a quality-flagged real-time RTK reference trajectory. A second collection extends the platform with a synchronized perception payload—stereo, omnidirectional and polarization cameras together with a rotating LiDAR—opening the dataset to vision- and polarization-aided approaches to the low-speed heading observability that the benchmark exposes. Baseline results for different estimators are reported to demonstrate usability. The data are organized in a RINEX and CSV format on a common GPS timescale, compatible with other benchmarks.