The goal of this project is the development of state-of-the-art methods and software for modeling and solving linear programs (LPs), mixed-integer programs (MIPs), mixed-integer nonlinear programs (MINLPs), and constraint integer programs (CIPs). At its center stands the development, maintenance, and distribution of the SCIP Optimization Suite (https://www.scipopt.org/).
SCIP is one of the fastest solvers for mixed-integer programming and mixed-integer nonlinear programming available in source code, and at the same time a flexible framework for constraint integer programming and branch-cut-and-price. It combines the strengths of integer programming — LP relaxations and cutting planes for strong dual bounds — with constraint programming techniques such as domain propagation for handling arbitrary, in particular nonlinear, constraints. As a framework, SCIP gives users full control of the solution process: nearly every algorithmic component can be extended or replaced through plugins for branching rules, cutting planes, primal heuristics, propagation, pricing, and Benders' decomposition. SCIP is free and open source under the Apache 2.0 license, and it is used in research, industry, and teaching worldwide.
Current research focuses on finding feasible solutions more efficiently, learning from infeasibilities, GPU-accelerated first-order methods for large-scale problems, parallelization of core solver components, and the integration of machine learning into the solving loop. This work is carried out within the Research Campus MODAL together with industry partners.
