TEX-CUP: The University of Texas Challenge for Urban Positioning

Low-Cost Inertial Aiding for Deep-Urban Tightly-Coupled Multi-Antenna Precise GNSS

“…3) A novel false fix detection and recovery technique that limits the degree to which an incorrectly-resolved integer ambiguity can corrupt the tightly-coupled CDGNSS estimator's state (Section IV-D). 4) Demonstration of state-of-the-art deep urban CDGNSS performance, achieving, by tightly coupling with consumergrade and industrial-grade inertial sensors, respectively, a 96.6% and 97.5% integer fix availability, and 12.0 cm and 10.1 cm overall (fix and float) 95th percentile horizontal positioning error on the publicly-available TEX-CUP urban positioning dataset [32] (Sections V-A to V-C).…”

“…The sensor platform described in this paper and used in the evaluation in Section V is the University of Texas Sensorium [32], a roof-mounted vehicular perception platform incorporating multiple grades of inertial sensor, two GNSS antennas (denoted primary and secondary), stereo cameras, and three automotive radars. Only the inertial sensors and GNSS antennas are used in this work.…”

“…The tightly-coupled CDGNSS estimator described in the foregoing sections was implemented in C++ as a new version of the PpEngine sensor fusion engine [4], and was experimentally evaluated against the publicly-available TEX-CUP urban positioning dataset. TEX-CUP comprises raw GNSS intermediatefrequency (IF) samples and inertial data collected on 9 and 12 May, 2019, using the University of Texas Sensorium vehicular perception research platform [32]. The dataset consists of a total of over 2 hours of driving in Austin, Texas in conditions ranging from light to dense urban; routes are shown in Fig.…”

Automotive-Radar-Based 50-cm Urban Positioning

“…V. EXPERIMENTAL RESULTS The radar-based urban positioning system was evaluated experimentally using the dataset described in [12], collected on Thursday, May 9, 2019 and Sunday, May 12, 2019 during approximately 1.5 h of driving each day in and around the urban center of Austin, TX. This section presents the evaluation results.…”

“…This paper also presents the first evaluation of low-cost automotive-grade radar-based urban positioning on the largescale dataset described in [12]. Data from automotive sensors are collected over two 1.5 h driving sessions through the urban center of Austin, TX on two separate days specifically chosen to provide variety in traffic and parking patterns.…”