Despite the decreasing cost of hardware and improving compentency of open source software large-scale robotics research is still extremely expensive. Few research groups are able to manage the costs of developing and maintaining a safe on-road data collection platform, regular calibration, data collection procedures and storing and processing the collected data. Especially in the era of data-driven problem solving approaches, it is critical that the research community is able to access vast quantities of real-world data for development, testing and validation of algorithms before deployment.

Modern, and ever increasingly autonomous, robots now need to see further, through fog, rain and snow, despite lens flare or when directly facing the sun. Millimetre-Wave radar holds the promise of consistent sensor observations in such conditions where other sensor modalities such as vision and LIDAR may fail.

In recent years we have benefited from more and more high quality datasets shared to the community, notably the KITTI, Cityscapes and Oxford RobotCar datasets. However, none of these offer Millimetre-Wave FMCW radar data and as we are fortunate enough to have collected a lot of data in this modality, this is something we wanted to share.

For now, we are releasing 32 traversals of the Oxford RobotCar Dataset route driven in January 2019 covering 280 km of driving.

The release includes:

  • 1 millimetre-Wave FMCW radar (Navtech CTS350-X)
  • 1 stereo camera (Point Grey Bumblebee XB3) and 3 monocular cameras (Point Grey Grasshopper 2)
  • 2 3D LIDARs (Velodyne HDL-32E)
  • 2 2D LIDARs (SICK LMS151)
  • 1 GPS / INS (NovAtel SPAN-CPT)

The processed data includes:

  • Visual odometry from the stereo camera
  • Optimised radar odometry

We may release more data in the future but for now we are very excited to share this data with the community and we intend that this dataset will encourage research in this interesting modality.

If you have any questions please don’t hesitate to get in touch:


This dataset collection wouldn’t have been possible without help from numerous people throughout The Oxford Robotics Institute for driving and co-driving and in particular the Engineering Team for hardware and software support during the trials.

We would also like to thank our partners at Navtech Radar, with whom this dataset release would not have been possible.