Prof. Dr. Ralf Borndörfer
Prof. Dr. Martin Skutella
Duration: 01.01.2012 - 31.12.2016
Technische Universität Berlin
/ Konrad-Zuse-Zentrum für Informationstechnik Berlin
The project "Multicriteria Optimisation" considers mathematical questions and discrete problems within the CRC 1026 "Sustainable Manufacturing". The three sustainability dimensions "economic", "environmental" and "social", respectively, are considered as different objective functions by the project A5. Hence, discrete problems and mathematical questions are modelled by a feasible space of solutions and several objectives which have to be optimised simultaneously. In contrast to the single-criteria case, it is generally not possible to find a solution which optimises all considered objectives simultaneously. Instead one has to deal with trade-offs. For example, the cheapest way to manufacture a certain amount of bicycle frames might not be the environmentally friendliest. A solution that can be improved in at least one objective without getting worse off in the other is called inefficient and will generally be neglected by a decision maker. Hence, only the efficient solutions are interesting from a decision maker's point of view. Besides the mathematical questions about the existence and number of efficient solutions and the algorithmic approaches of how to compute them, the project A5 is also concerned with the modelling of quantitative problems within the CRC 1026. With respect to models the focus and expertise is on mixed integer programming.
We have developed PolySCIP, an open-source and freely available solver which aims at solving multicriteria mixed integer programs with an arbitrary number of objectives. With respect to scenario analysis two tools, tech-con and field-con, were implemented. Exemplary applications like the optimization of process chains for bicycle frame manufacturing, the selection of sustainable welding processes and design decision support are documented.