Datasets
YouTube
Robot and Sensors
We utilize a quadrupedal robot as a base platform to collect the dataset and set up the compact sensor configuration.
To capture visual measurements with different wavelengths, the RGB and thermal cameras are placed to form a forward-looking sensor system.
The terrain-looking sensor system uses an RGB-D camera to obtain sufficient ground information.
In addition, LiDAR enables the recognition of the surrounding environment regardless of angle.
Finally, GPS, IMU, and built-in odometry are responsible for the robot's Navigation.
Data Format
We provide our datasets in rosbag format. Each perceptual sensor and motion sensor are represented in this table.
Perceptual SensorsÂ
From left to right, sub map from LiDAR, and sample images of ground-facing, forward viewing, thermal, and pointcloud from depth camera are shown.
Motion Sensors
Graphs indicate the roll and pitch value in angular velocity from IMU in [rad/s].
PCD
We also provide LiDAR maps for each sequences.