HyperVariables

General variable and Lie group containers with extensive Eigen interoperability for non-linear optimizations.
Report Issues or Request Features

About

HyperVariables comprises low-level containers used in HyperSLAM to represent parameters within non-linear optimizations. All implemented variables are fully interoperable with Eigen and, as such, are mappable from raw memory locations. All variables inherit from an abstract base, which allows storing non-homogeneous variables in a simple and unified manner. At present, three different distortion models along with two common Lie groups and other common variables are implemented. Where adequate, we also provide analytic Jacobians for the implemented operations. If you use this repository, please cite it as below.

@article{RAL2022Hug,
    author={Hug, David and B\"anninger, Philipp and Alzugaray, Ignacio and Chli, Margarita},
    journal={IEEE Robotics and Automation Letters},
    title={Continuous-Time Stereo-Inertial Odometry},
    year={2022},
    volume={7},
    number={3},
    pages={6455-6462},
    doi={10.1109/LRA.2022.3173705}
}

Installation

HyperVariables depends on the Eigen, Google Logging and Google Test libraries and uses features from the C++20 standard (see link to update gcc and g++ alternatives). The setup process itself (without additional compile flags) is as follows:

# Clone repository.
git clone https://github.com/VIS4ROB-lab/HyperVariables.git && cd HyperVariables/

# Run installation.
chmod +x install.sh
sudo install.sh

# Build repository.
mkdir build && cd build
cmake ..
make

Literature

1. Continuous-Time Stereo-Inertial Odometry, Hug et al. (2022)
2. HyperSLAM: A Generic and Modular Approach to Sensor Fusion and Simultaneous
   Localization And Mapping in Continuous-Time, Hug and Chli (2020)
3. A Micro Lie Theory for State Estimation in Robotics, Solà et al. (2018)
4. A Generic Camera Model and Calibration Method for Conventional,
   Wide-Angle, and Fish-Eye Lenses, Kannala and Brandt (2006)
5. Single View Point Omnidirectional Camera Calibration from Planar Grids, Mei and Rives (2007)

Known Issues

1. Tests covering the (equidistant) distortion models occasionally fail due to surpassing the numeric tolerances.

Updates

17.06.22 Initial release of HyperVariables.

Contact

Admin - David Hug, Leonhardstrasse 21, 8092 Zürich, ETH Zürich, Switzerland
Maintainer - Philipp Bänninger, Leonhardstrasse 21, 8092 Zürich, ETH Zürich, Switzerland
Maintainer - Ignacio Alzugaray, Leonhardstrasse 21, 8092 Zürich, ETH Zürich, Switzerland

License

HyperVariables are distributed under the BSD-3-Clause License.