Software

GitLab Repository des Instituts für Mathematische Optimierung

Eine Auswahl an Softwarepaketen finden Sie auch auf unserem GitLab Repository

ISPoc

ISPoc (Intrusive Spectral Projection for optimal control) is a framework for transforming uncertain optimal control problems into deterministic surrogates, which can be solved with a direct method for optimal control (MUSCOD-II). It contains a symbolic implementation of the intrusive spectral projections method (Polynomial Chaos expansion).

• A preliminary version of the ISPoc soure code is available for download.

MUSCOD-II

(with Simulation & Optimization group at IWR Heidelberg)

Our in-house mixed-integer optimal control software package MUSCOD-II now has an AMPL interface that allows to model DAE-constrained optimal control problems in AMPL.

• An interface to a small subset of MUSCOD-II is available online through the NEOS Server for Optimization.

netmic

A python package that computes Irreducible Infeasible Subsystem Arc Covers in Flow Networks; included is a generator for infeasible network flow problems.

• netmic 0.1
• Documentation (tar.gz)

pySLEQP

pySLEQP is a prototypical implementation of a nonlinear programming solver employing a sequential linear equality constrained quadratic programming method.

• The pySLEQP soure code is available for download,
• as is a preprint describing the method and implementation.

qpOASES

(with H.J. Ferreau at ABB Baden/Switzerland, A. Potschka at IWR Heidelberg)

qpOASES is a primal-dual active set code for dense and sparse parametric QPs. It is well suited for hot-starting when solving a sequence of related QPs, such as in model predictive control or sequential quadratic programming. qpOASES is now hosted by the COIN-OR foundation and is available for free under the GNU LGPL license.

• The qpOASES source code is available from a COIN-OR git repository on GitHub.

SCARP

SCARP is a code computing solution to mixed-integer optimal control problems. It computes solutions by rounding solutions of fractional relaxations while taking switching costs into account, enabling the penalization of chattering, extending previous modeling capabilities.

• The SCARP source code can be obtained from GitHub.

SCIP

SCIP is a Branch-and-Cut framework and also contains a fast integer optimization solver. SCIP is developed together with the Zuse Institute Berlin and the Universitity of Erlangen-Nürnberg, Chair of EDOM.

• SCIP project page
• SCIP is free for academic purposes - see  license information for more information.

SoPlex

SoPlex is a fast linear optimization solver. SoPlex is developed together with the Zuse Institute Berlin.

• SoPlex project page
• SoPlex is free for academic purposes - see license information for more information.

TACO

(with S. Leyffer at Argonne National Labs)

TACO is the Toolkit for AMPL Control Optimization that drives MUSCOD-II on NEOS. It defines some add-ons to the AMPL modeling language that allow the elegant formulation of ODE/DAE optimal control problems in AMPL. Developers of optimal control problem solvers may want to take a look at the TACO source code to interface their solvers with AMPL.

• The TACO source code is available for download.

trlib

(with Felix Lenders, Andreas Potschka at IWR Heidelberg)

This is a library that provides various methods related to solving the trust region subproblem in an iterative way. The principal approach follows the GLTR method by N. Gould.

• The trlib source code is available on GitHub.
• The trlib preprint has all the details.

trlib is now part of SciPy and can be called via scipy.optimize.minimize.