A New Approach to 3D ICP Covariance Estimation for Mobile Robotics [paper]

Overview

In mobile robotics, scan matching of point clouds using Iterative Closest Point (ICP) allows estimating sensor displacements. It may prove important to assess the associated uncertainty about the obtained rigid transformation, especially for sensor fusion purposes. We propose a novel approach to 3D ICP covariance computation that accounts for all the sources of errors as listed in Censi's pioneering work, namely wrong convergence, underconstrained situations, and sensor noise. Our approach builds on two facts. First, ICP is not a standard sensor: owing to wrong convergence the concept of ICP covariance per se is actually meaningless, as the dispersion in the ICP outputs may largely depend on the accuracy of the initialization, and is thus inherently related to the prior uncertainty on the displacement. We capture this using the unscented transform, which also reflects correlations between initial and final uncertainties. Then, assuming white sensor noise leads to overoptimism: ICP is biased, owing to e.g. calibration biases, which we account for. Our solution is tested on publicly available real data ranging from structured to unstructured environments, where our algorithm predicts consistent results with actual uncertainty, and compares very favorably to previous methods. We finally demonstrate the benefits of our method for pose-graph localization, where our approach improves accuracy and robustness of the estimates.

Code

This repo contains the code for reproducing the results of this paper. The code is based on Python and has been tested under Python 3.5 on a Ubuntu 16.04 machine. ICP algorithm is called throught our modified version of the libpointmatcher library.

Installation & Prerequies

1. Clone this repo

git clone https://github.com/CAOR-MINES-ParisTech/3d-icp-cov.git
mkdir data
mkdir results

2. Install the following required Python packages, matplotlib, numpy, scipy, alphashape, e.g. with the pip command

pip install matplotlib numpy scipy alphashape

3. Clone our fork of the libpointmatcher library and build them the library

cd 3d-icp-cov
git clone https://github.com/CAOR-MINES-ParisTech/libpointmatcher.git
cd libpointmatcher
mkdir build && cd build
cmake ..
make
sudo make install
cd ../..

4. Download the Challenging data sets for point cloud registration algorithms. Extract the zip file of each sequence corresponding to the point clouds in base frame in the data folder

Get started

1. Modify paths and parameters in the class Param at the end of python/utils.py

2. Launch the main file

   cd python
   python3 main.py
   

   Be patient, it will reproduce the results of the paper.

Paper

The paper A New Approach to 3D ICP Covariance Estimation for Mobile Robotics, M. Brossard, S. Bonnabel and A. Barrau. 2019, relative to this repo is available at this url.

Citation

If you use this code in your research, please cite:

@article{brossard2019anew,
  author = {Martin Brossard and Silv\`ere Bonnabel and Axel Barrau},
  title = {{A New Approach to 3D ICP Covariance Estimation for Mobile Robotics}},
  year = {2019}
}

If you use the original libpointmatcher library in your research, please cite the original paper:

@article{Pomerleau12comp,
	author = {Pomerleau, Fran{\c c}ois and Colas, Francis and Siegwart, Roland and Magnenat, St{\'e}phane},
	title = {{Comparing ICP Variants on Real-World Data Sets}},
	journal = {Autonomous Robots},
	year = {2013},
	volume = {34},
	number = {3},
	pages = {133--148},
	month = feb
}

Authors

Martin Brossard^, Silvère Bonnabel^ and Axel Barrau°

^MINES ParisTech, PSL Research University, Centre for Robotics, 60 Boulevard Saint-Michel, 75006 Paris, France

°Safran Tech, Groupe Safran, Rue des Jeunes Bois-Châteaufort, 78772, Magny Les Hameaux Cedex, France