Learning Wheel Odometry and IMU Errors for Localization
Martin Brossard and Silvère Bonnabel
This repo containt the Python code for reproducing the results of the paper Learning Wheel Odometry and IMU Errors for Localization. Please follow the links to read the paper.
Installation & Pre-Requisites
- Install the master version of PyTorch, the development version of pyro, liegroups and progressbar. Remaining packages are standard Python packages. All our code was running with Python 3.5.
- Download data from one or two datasets (see below)
- Clone the current repo
git clone https://github.com/Center-for-Robotics-MINES-ParisTech/gpkf
University of Michigan North Campus Long-Term Vision and LiDAR Dataset
The Segway dataset is described in the following paper:
- Nicholas Carlevaris-Bianco, Arash K. Ushani, and Ryan M. Eustice, University of Michigan North Campus Long-Term Vision and Lidar Dataset, International Journal of Robotics Research, 2016.
Dataset can be downloaded following this link and extracted in data/nclt
.
- Training data: first 19 sequences
- Cross-validation data:
2012-10-28
,2012-11-04
,2012-11-16
,2012-11-17
- Testing data:
2012-12-01
,2013-01-10
,2013-02-23
,2013-04-05
Complex Urban LiDAR Data Set
The car dataset is based on the paper
- Jinyong Jeong, Younggun Cho, Young-Sik Shin, Hyunchul Roh, Ayoung Kim, Complex Urban LiDAR Data Set, 2018.
Dataset can be downloaded following this link and extracted in data/kaist
.
- Training data:
urban00
tourban11
andcampus00
- Cross-validation data:
urban12
,urban13
,urban14
- Testing data:
urban15
,urban16
Training and Testing
- Modify setting and parameters if nessesary in
main_nclt.py
ormain_kaist.py
- Run
main_nclt.py
ormain_kaist.py
Citing the paper
If you find this code useful for your research, please consider citing the following paper:
@unpublished{brossard2018Learning,
Title = {Learning Wheel Odometry and IMU Errors for Localization},
Author = {Brossard, Martin and and Bonnabel Silvère},
Year = {2019}
}
License
For academic usage, the code is released under the permissive MIT license.
Acknowledgements
We thank the authors of the University of Michigan North Campus Long-Term Vision and LiDAR Dataset and especially Arash \textsc{Ushani} for sharing their wheel encoder data log.