XJTU Campus Dataset: RTK Ground Truth for GNSS-Visual-Inertial Odometry (7 Sequences)

The XJTU Campus Dataset comprises seven visual-inertial-GNSS sequences collected at the Xi'an Jiaotong University campus and surrounding areas for GNSS-visual-inertial odometry evaluation. This release accompanies the Mechatronics paper on R-GVINS. This record includes:- RTK Fixed ground-truth trajectories for all seven sequences (TUM format, local ENU frame)- Sensor calibration parameters (Intel RealSense D435i camera/IMU, u-blox ZED-F9P GNSS)- Sequence metadata (duration, path length, RTK Fixed rate)- RTK trajectory overview maps Collection summary: 186.6 minutes (~3.1 hours), ~10.6 km total path length. Environments include campus (day/night), dormitory (day/night), indoor-outdoor library transition, and sports field. Ground truth is derived from RTK fixed solutions (centimeter-level; RMS receiver-reported accuracy: 1.84 cm 2D, 3.26 cm 3D across all sequences). Raw ROS bag files (~290 GB) are not included in this record due to file size. Processed ground-truth trajectories and calibration are provided to support reproducibility of reported evaluation metrics. Related software: R-GVINS (GitHub URL to be added upon code release).