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

1. 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.
2. Download data from one or two datasets (see below)
3. 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 to urban11 and campus00
• Cross-validation data: urban12, urban13, urban14
• Testing data: urban15, urban16

Training and Testing

1. Modify setting and parameters if nessesary in main_nclt.py or main_kaist.py
2. Run main_nclt.py or main_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@unpublished{brossard2018Learning,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.