27 October 2016
Autonomous vehicle research is very much dependent on high volumes of real-world data for development, testing and validation of algorithms before deployment on public roads. The University of Oxford just did every autonomous car developer a solid by releasing 1000km (600+ miles) driverless road data of their ‘Robotcar‘.
Few research groups can manage the costs of developing and maintaining a suitable autonomous vehicle platform, let alone regularly calibrating or storing and processing the collected data. The University of Oxford does believe these resources should be available to anyway. To put everyone a step in the right direction the university set a new benchmark: the Oxford RobotCar Dataset.
Over the period of November 2014 to December 2015 the driverless team traversed a 10km route through central Oxford twice a week on average in the Oxford RobotCar platform, an autonomous Nissan LEAF. This resulted in approximately 1000km of recorded driving with over 20 million images collected from 6 cameras mounted to the vehicle, along with LIDAR, GPS and INS ground truth. The University of Oxford robotcar data was collected in all weather conditions, including heavy rain, at night time, in direct sunlight and snow, or with hindering road and building works. By frequently traversing the same route over the period of a year the university enabled investigating long-term localisation and mapping for autonomous vehicles in real-world, dynamic urban environments.
Before the University of Oxford robotcar data, a number of vision-based autonomous driving datasets had already been released, notably the KITTI dataset and the Cityscapes dataset. But neither of these datasets addressed the challenges of long-term autonomy, localisation in the same environment under significantly different conditions and mapping in the presence of structural change over time.
The data can be found here: http://robotcar-dataset.robots.ox.ac.uk/