Public datasets‎ > ‎

Obstacle Traversal Dataset 2017

This dataset was created in laboratories of the Czech Technical University in Prague, in outdoor areas near Prague and in an industrial training plant RDM in Rotterdam. Sensor data are recorded using Robot Operating System (ROS, version Indigo) as a bag file.

Laboratory with HTC Vive tracker Laboratory with HTC Vive tracker 
Outdoor stairway RDM training plant

There are several ROS topics recorded: laser scanner, odometry, IMU, force sensors in track flippers and many more. In order to keep bag files reasonably large, most of the experiments do not contain camera images. However, 3D laser scans give a good idea about each particular experiment. 

Geometry of the robot is stored in /tf topic together with the basic 6-DOF odometry - transformation from the /odom (odometry) frame to the /base_link (Body or Robot) frame. It was evaluated on-board during the experiments. Most of indoor experiments contain position reference provided by a 3D tracker using the HTC Vive virtual reality headset system (link). This reference is expressed as a transformation from the /world frame to the /base_link_htc_tracked frame. This transformation is disconnected from the from the robot transformation tree because of limitation laid by the TF framework. The tracker is calibrated in order to align the /base_link_htc_tracked and the /base_link frames.

The dataset is organized by the type of tested motion or environment. Each folder contains a text file with rosbag info that lists contents of all the bag files in the folder, their size and length:

Short ramp

Long ramp

Single pallet straight forward

Ramp and steps Short obstacle

Staircase segment

Indoor staircase

Three steps

Obstacle course