Consensus NDT Odometry

Authors

Ashwin Vivek Kanhere and Grace Xingxin Gao, Stanford University

Link to paper: LiDAR SLAM Utilizing Normal DistributionTransform and Measurement Consensus

Implementation of LiDAR odometry using the Normal Distributions Transform (NDT) and measurement consensus. Measurement consensus is used to remove potentially faulty point clouds from the odometry optimization resulting in a speed-up when compared to an equivalent naive NDT odometry optimization.

Table of Contents

• Consensus NDT Odometry
  • Authors
  • Table of Contents
  • Prerequisites
  • Getting Started
    • Installation
  • Usage
  • License
  • Contact
  • Citation
  • Acknowledgements

Prerequisites

• Python 3.6
• Python modules listed below are required to run the code. We strongly recommend using 'requirements.txt' to install these modules.
  • Numpy
  • Scipy
  • Transforms3D
  • Numpy-Indexed
  • Points Processing Toolkit (pptk): PPTK requires Python 3.6 or lower and does not work on recent Python 3.xx versions
  • Matplotlib
  • scikit-learn
  • PyKITTI
  • OpenCV for Python

Getting Started

Installation

1. Clone the repository.

git clone https://github.com/kanhereashwin/ion-gnss-19.git

2. Install packages. We recommend using Anaconda for package management.

pip install -r requirements.txt

Usage

1. Update the location of datasets in 'data_utils.py' for both the KITTI and UIUC Point cloud datasets.
2. Run 'main_implementation.py' to generate comparisons between naive NDT and Consensus NDT implementations for position, rotation accuracies and time taken to compute odometry.

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Ashwin Vivek Kanhere - akanhere {at} stanford.edu Navigation and Autonomous Vehicles Lab Department of Aeronautics and Astronautics Stanford University

Project Link: https://github.com/kanhereashwin/ion-gnss-19

Citation

If you use this code for your research, please cite our paper:

@inproceedings{kanhere2019consensus,
  title={LiDAR SLAM Utilizing Normal Distribution Transform and Measurement Consensus},
  author={Kanhere, Ashwin Vivek and Gao, Grace Xingxin},
  booktitle={32nd International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS},
  pages={2228 - 2240},
  year={2019}
}

Acknowledgements

Code borrows the ICP implementation from https://github.com/ClayFlannigan/icp/blob/master/icp.py We also acknowledge Siddarth Tanwar and Shubh Gupta for their comments and reviews.