This page contains the data and information for the LiDAR Inertial dataset in environments with large sections of degenerate geometry that was presented in the ICRA 2024 publication:

Patrick Pfreundschuh, Helen Oleynikova, Cesar Cadena, Roland Siegwart, Olov Andersson, “COIN-LIO: Complementary Intensity-Augmented LiDAR Inertial Odometry”, IEEE International Conference on Robotics and Automation (ICRA), 2024.

Datasets

The dataset was recorded using an Ouster OS0 128 (Rev D) LiDAR at 10 Hz with 1024×128 points per revolution. The IMU data is recorded at 100 Hz using the internal IMU of the LiDAR.

The ground-truth position was recorded from a Leica MS60 Total using a Prism mounted on top of the LiDAR and time-synced accordingly. In the recorded sequences, the sensor is carried around on a handheld stick. For each environment, we provide one sequence smooth (walking, slow motions) and one dynamic (running, aggressive motions), indicated by either “s” (smooth) or “d” (dynamic) at the end of the filename, e.g. 2023-08-08-17-50-31-tunnel_d.bag

The sequences are provided as rosbags. Ground truth files are provided as CSV in TUM format.

IMU to prism extrinsics were calculated using CAD and can be found here.

Exemplary images of each environment, as well as coordinates of the starting position of each sequence (manually annotated estimate):

Intersection:

Location

Runway:

Location

Field:

Location

Tunnel:

Location

Katzensee:

Location