Awesome

Copra

Copra (Control & preview algorithms) is a C++ library implementing linear model predictive control. It relies on quadratic programming (QP) solvers. Python bindings are available.

main image @Stephane Caron

This work was originally made by Vincent Samy and extended by Niels Dehio.

Copra is licensed under the BSD-2-Clause. However, its default QP solver (eigen-quadprog) is licensed under the LGPL-2 and this cannot be changed. Please be aware of the related restrictions if you plan to work with copra. At a later date, we might switch copra default QP solver to one with a less restrictive license.

Installation

Copra should compiles and is tested on Linux, macOS and Windows.

Ubuntu LTS (16.04, 18.04, 20.04)

You must first setup our package mirror:

curl -1sLf \
  'https://dl.cloudsmith.io/public/mc-rtc/stable/setup.deb.sh' \
  | sudo -E bash

You can also choose the head mirror which will have the latest version of this package:

curl -1sLf \
  'https://dl.cloudsmith.io/public/mc-rtc/stable/setup.deb.sh' \
  | sudo -E bash

You can then install the package:

sudo apt install libcopra-dev

Dependencies

C++ compiler with C++14 support (see below for C++11 support)
CMake >= 2.8
Doxygen: to generate documentation (optional)
Eigen >= 3.2
eigen-quadprog

Optional

Boost for Python bindings (>= 1.58) and unit tests
GUROBI >= 4.0: optional QP solver
eigen-gurobi: bindings for GUROBI
eigen-qld: optional QP solver
eigen-osqp: optional QP solver
pygen-converter: for python bindings

Building from source on Linux

The library is written in c++14. Compiler that can not support it won't be able to compile it.

git clone --recursive git@github.com:jrl-umi3218/copra.git
cd Copra
mkdir build && cd build
cmake ..
ccmake .  # configure e.g. PYTHON_BINDING or BUILD_TESTING
make -j4
sudo make install

C+11 compatible code

A c++11 compatible is available on a specific branch. Although it exists, it may not be updated as often as the master branch. The branch is called c++11-compatible. To use it, before compilation and after cloning the repo, do:

git checkout c++11-compatible

Then follow the steps in the section just above.

Testing

C++ tests will be compiled in your build folder if you enabled the BUILD_TESTING option:

cd build/tests
./TestSolvers --log_level=all
./TestLMPC --log_level=all

Once Python bindings are installed, you can check them with:

cd binding/python/tests
python TestMPController.py

Documentation

Doxygen documentation is available online

You can also check out unit tests, as well as the following two examples:

Reference

Linear model predictive control including optimization of the initial state employing COPRA has been described in

@unpublished{Dehio2021ICRA,
  title = {Safe Impacts with Soft Contacts Based on Learned Deformations},
  author = {Dehio, Niels and Kheddar, Abderrahmane},
  booktitle={IEEE Int. Conf. on Robotics and Automation},
  pdf = {https://hal.archives-ouvertes.fr/hal-02973947/document},
  url = {https://hal.archives-ouvertes.fr/hal-02973947},
  year = {2021}
}

Writing code takes time. If this implementation is useful for your research, please cite the related publication.