LCQPow is a open-source solver for Quadratic Programs with Complementarity Constraints. The approach is based on a standard penalty homotopy reformulated using sequential convex programming. The convex sequence derives from linearizing the (necessarily) nonconvex penalty function. This leads to a constant objective Hessian matrix throughout all iterates, and thus enables us to solve the linear complementarity quadratic program with a single factorization of the KKT matrix (by using qpOASES).
The entire strategy is presented in detail in this paper.
- Build process is currently only tested on Ubuntu 20.04
- CMake version >= 3.20.3
- All dependencies are linked automatically. These are
- Clone the repository, and recursively initialize the submodules
$ git clone https://github.com/hallfjonas/LCQPow.git
$ cd LCQPow
$ git submodule update --init --recursive
- Configure and build the project
$ mkdir build
$ cd build
$ cmake ..
$ make
- To test the build you can run the test examples
$ bin/examples/warm_up
$ bin/examples/warm_up_sparse
$ bin/examples/warm_up_osqp
$ bin/examples/OptimizeOnCircle
The MATLAB interface is built automatically if matlab is successfully detected by CMake. Make sure that your linker can locate the created libraries, e.g. by exporting the library path in the same shell as the one you call matlab in:
$ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:<LCQPow-dir>/build/lib
$ matlab
where <LCQPow-dir>
represents the base directory of your local LCQPow repository.
Whenever using the MATLAB interface make sure that MATLAB is able to locate the interface by executing
addpath("<LCQPow-dir>/build/lib")
Navigate your MATLAB editor to the directory <LCQPow-dir>/interfaces/matlab/examples
and play with any of the provided codes.
Type help LCQPow
in order to obtain the documentation of our MATLAB interface.
The file LICENSE contains a copy of the GNU Lesser General Public License (v2.1). Please read it carefully before using LCQPow!
If you have got questions, remarks or comments on LCQPow, it is strongly encouraged to report them by creating a new issue on this github page.
Finally, you may contact the main author directly: Jonas Hall, [email protected]
Also bug reports, source code enhancements or success stories are most welcome!