Real-time Kinematic Ground Truth

We are pleased to release the reference data towards a planned challenging long-term localisation and mapping Oxford RobotCar Long-Term Autonomy Benchmark. Using post-processed raw GPS, IMU, and static GNSS base station recordings, we have produced a globally-consistent centimetre-accurate ground truth for 72 traversals across the year-long duration of the dataset. We exclude from this release the six RobotCar Seasons runs.

For more information describing the ground truth and if you use this reference data in your research, please cite the following paper:

W. Maddern, G. Pascoe, M. Gadd, D. Barnes, B. Yeomans, and P. Newman, “Real-time Kinematic Ground Truth for the Oxford RobotCar Dataset”, in arXiv preprint arXiv: 2002.10152, 2020. [Bibtex][PDF]

@article{RCDRTKArXiv,
  author = {Will Maddern and Geoffrey Pascoe and Matthew Gadd and Dan Barnes and Brian Yeomans and Paul Newman},
  title = {Real-time Kinematic Ground Truth for the Oxford RobotCar Dataset},
  journal = {arXiv preprint arXiv: 2002.10152},
  url = {https://arxiv.org/pdf/2002.10152},
  pdf = {https://arxiv.org/pdf/2002.10152.pdf},
  year = {2020}
}

The RTK ground truth files are prepared in a CSV format compatible with those described in Documentation and are readable by the updated robotcar-dataset-sdk. The date string for each CSV corresponds to those shown on Datasets. The files are collected in a single archive (approximately 91MB compressed) which can be downloaded at the following link: