The Oxford RobotCar dataset

Long-term localisation and mapping in changing environments is a significant challenge in the race to develop autonomous vehicles. Critical to development is the availability of large quantities of real-world data, which is used to develop, test, and validate algorithms ahead of successful deployment. Currently, available datasets do not address the challenges of long-term autonomy.

The University of Oxford Mobile Robotics Group has recorded over 20TB of image, LIDAR, and GPS data by repeatedly traversing a route in central Oxford in the Oxford RobotCar. This unique dataset, captured over one year, features a large range of variation in scene appearance and structure. The dataset is available to license for commercial use from Oxford University Innovation and will enable the development and testing of long-term mobile autonomy solutions.

The challenge of long-term localisation

Autonomous vehicle research is critically dependent on vast quantities of real-world data for development, testing, and validation of algorithms prior to deployment. A number of vision-based autonomous vehicle datasets have been released, however, they do not address the main challenges of mobile autonomy.
These challenges include localisation in the same environment under significantly different conditions and mapping in the presence of structural change over time.

The Oxford RobotCar dataset

Researchers from the globally renowned Oxford Robotics Institute have collected more than 20TB images, LiDAR and GPS data by repeatedly traversing a route in central Oxford in the Oxford RobotCar. The data was collected over a period of one year and represents over 1000km of recorded driving.

The resulting unique dataset captures a large range of variation in scene appearance and structure due to:

  • illumination
  • weather
  • dynamic objects
  • seasonal effects
  • construction

The raw recordings from all of the available sensors are included alongside a full set of intrinsic and extrinsic sensor calibrations. MATLAB development tools are additionally included for accessing and manipulating the raw sensor data.

Commercialisation

The Oxford RobotCar dataset offers a unique opportunity for the commercial development and testing of systems capable of long-term localisation in dynamic environments. The dataset is available to license from Oxford University Innovation.

For further information on data collection and utility, please visit http://robotcar-dataset.robots.ox.ac.uk/images/robotcar_ijrr.pdf

Request more information
about this technology
Search

Ready to get in touch?

Contact Us
Search
© Oxford University Innovation