Stereo-Camera Main Platform Calibration

The dataset is collected onboard a compact handheld device that includes:

  • Intel NUC8i7BEH
  • Ouster OS-1 (Gen 1) 64 beam LIDAR with embedded IMU
  • RealSense D435i stereo camera with embedded IMU
  • 8000 mAh battery pack and onboard power electronics

The Intel Realsense D435i is a commodity-grade stereo-inertial camera while the Ouster OS-1 LiDAR has 64 beams, which provides much denser data than many other LiDAR datasets. Both sensors have become commonly used in mobile robotics in the last 2 years, for example the ongoing DARPA Subterranean Challenge.

Sensor Type Rate [Hz] Characteristics
Ouster OS-1-64 LiDAR 10 64 beams, 120 m max range, 33.2° nominal vertical FoV
ICM-20948 IMU 100 Ouster embedded
Intel RealSense D435i Stereo Camera 30 Global shutter (IR), 848 × 480
Bosch BMI055 IMU 650 (400 gyro + 250 accel) RealSense embedded

The sensors are rigidly attached to a precisely 3D-printed base. The base has a modular design which allows to install up to 4 stereo cameras facing in opposite directions with variable downward facing directions.

handheld_device

The current design has one RealSense with a 0 deg pitch angle.

To distinguish the sensor frames, we use the following abbreviations:

  • OS_I : The IMU coordinate system in the LiDAR.
  • OS_L : The LiDAR coordinate system with respect to which the point clouds are read
  • RS C1: The left camera optical frame. Note: The base frame is defined to be coincident with the left camera frame; however, it follows the robotic frame (x-forward, y-left, z-up).
  • RS_I : The IMU coordinate system in the stereo setup.
  • RS C2: The right camera optical frame

A URDF of the platform is publicly available at this repo https://github.com/ori-drs/frontier_description, branch newer-college-dataset. To visualise the model you can run:

roslaunch frontier_description view_urdf.launch