To path indicator To navigation To quick access To footer with other services
To content
Logo Chair of Modeling and Simulation
Department of Computer Science
Studenten im Sommer auf dem Campus Nord, rote Zahnräder im Hintergrund © Roland Baege​/​TU Dortmund

Graduiertenkolleg 1855 "Discrete Optimization of Technical Systems under Uncertainty".

The optimization of technical systems is the central topic of the Research Training Group. Real optimization problems in this context are often characterized by the fact that possible values of the decision variables are chosen from a discrete set, there is a certain uncertainty with respect to various internal and external factors, and the solutions determined by algorithms have to be implemented by human operators and also have to be answered for. This results in requirements for practically usable solutions that go beyond the algorithmic solution of optimization problems with deterministic objective functions and take into account the inherent uncertainty as well as humans as part of the decision process.

The Research Training Group combines methodological sciences with applied sciences and cognitive psychology in order to arrive at practically usable solutions based on sound theory. The subject matter is strongly interdisciplinary and includes three concrete application areas from the engineering sciences. The methodology of discrete optimization under uncertainty - including the implementation of optimization methods in algorithms - is the central aspect of the research training group, which is complemented by the interaction with the human operator of an engineering system. Discrete optimization under uncertainty is a current research topic with immense application potential. The methodological and algorithmic problems involved are very challenging; their solution is of great scientific interest and offers a long-term perspective. Internationally, many groups are working on individual aspects of the topic - but the cross-sectional topic chosen in the Research Training Group represents a unique selling point.

The development and operation of technical systems like production systems, logistics networks or large IT systems has to be based on a large number of design and configuration decisions to meet the performance requirements with a limited amount of resources and costs. Necessary decisions are often based on the solution of optimization  problems  with  discrete  or  mixed  discrete-continuous  parameters  describing the available alternatives.

Optimization problems of this kind are hard to solve as the number of available solutions exponentially increases with the number of decisions between discrete alternatives due to the “combinatorial explosion”. Most practical problems are simplified significantly to allow an algorithmic solution. Furthermore, in practice, decisions often have to be made with incomplete knowledge. The resulting uncertainty is usually not considered in existing optimization approaches even if this may result in considerable differences between the computed and real solution of the optimization problem. In some cases computed solutions may not even be feasible in practice.  

Another yet not deeply considered aspect of the optimization of technical systems is the role of people in the decision process. Mathematical methods and algorithms may compute optimal parameter values but the final solution must be accepted by a person and must be translated into concrete plans and instructions. To increase the applicability of optimization methods in practice, people must be regarded as part of the decision process. This implies that the process of optimization and result representation  must take into account the requirements of users.

The topic of the graduate school is optimization under uncertainty with the incorporation of people in the optimization process. Application scenarios that will be considered occur the areas of logistics, chemical production systems and IT systems.

Topics of the graduate school are interdisciplinary since the school combines research on methods from optimization, algorithms, statistics, applied science and psychology.  As doctoral  candidates  with  different  backgrounds  are  expected  to  join  the  graduate school they will attend basic courses to obtain a common foundation for their research and they will participate in specific courses to obtain a deep knowledge in the area of their doctoral research projects.

The optimization of technical systems is the central topic of the Research Training Group.  Real optimization problems in this context are often characterized by the fact that possible values of the decision variables are chosen from a discrete set, there is a certain uncertainty with respect to various internal and external factors, and the solutions determined by algorithms have to be implemented by human operators and also have to be answered for.  This results in requirements for practically usable solutions that go beyond the algorithmic solution of optimization problems with deterministic objective functions and take into account the inherent uncertainty as well as humans as part of the decision process.

The Research Training Group combines methodological sciences with applied sciences and cognitive psychology in order to arrive at practically usable solutions based on sound theory.  The subject matter is strongly interdisciplinary and includes three concrete application areas from the engineering sciences. The methodology of discrete optimization under uncertainty - including the implementation of optimization methods in algorithms - is the central aspect of the research training group, which is complemented by the interaction with the human operator of an engineering system. Discrete optimization under uncertainty is a current research topic with immense application potential. The methodological and algorithmic problems involved are very challenging; their solution is of great scientific interest and offers a long-term perspective.  Internationally, many groups are working on individual aspects of the topic - but the cross-sectional topic chosen in the Research Training Group represents a unique selling point.

The research program takes interdisciplinarity into account by the fact that many doctoral topics lie at the interface between different disciplines and have been jointly formulated by researchers from different fields. Discrete optimization under uncertainty is addressed in several ways. In particular, the existing methods to incorporate uncertainty into discrete optimization models are investigated in great breadth. The PhD projects from the three application areas use the algorithms developed in the methods area, extend them in an application-specific way, and thus also define new objectives for method development.  On the one hand, work from the area of user interaction complements the optimization algorithms by incorporating human decisions into the optimization process, thus making use of knowledge that cannot be formalized or is insufficiently formalized and human expert knowledge.  On the other hand, this area deals with more classical topics of human-machine interaction - namely the representation of uncertainty and the development of problem-adapted user interfaces.  The different approaches become practical by integrating algorithms, applications and user interfaces into a research environment. With the help of the research environment, experiments can be conducted to compare and evaluate algorithms, application problems can be analyzed with new algorithms, and finally, it is possible to empirically evaluate user interfaces with the help of test subjects.

The technical breadth of the research program can only be successfully implemented in the dissertation projects if the doctoral students work closely together beyond the boundaries of their own disciplines. The prerequisite for such cooperation is a common understanding of the problem and a common terminology, which must be conveyed through the qualification concept of the graduate school. The qualification program includes elements that ensure that all doctoral students have a basic knowledge of the overall topic. In addition, there will be individual parts that optimally prepare you for your own field of work and its immediate environment. For example, doctoral students who work on specific methodological problems of optimization under uncertainty gain a deeper insight into at least one specific area of application and deal with the integration of the methods they have developed into a research environment. The technical courses are supplemented by courses on interdisciplinary key qualifications. Doctoral students are introduced to international research at an early stage through an international guest scientist program, a stay abroad and broad support in the preparation of contributions for international conferences. The supervision concept with two supervisors from different areas supplements the study program in the transfer of interdisciplinary breadth.

Members

  • Nicole Kusmierz  
    • Phone: (+49)231 755-2117  
    • E-Mail: nicole.kusmierz (at) cs.tu-dortmund.de
  • Dr. Paolo Campigotto  
  • Dr. Moritz Mühlenthaler  
    • Phone: (+49)231 755-7226 
    • E-Mail: moritz.muehlenthaler (at) math.tu-dortmund.de
  • Dr. Dimitri Scheftelowitsch  
    • Phone: (+49)231 755-5855  
    • E-Mail: dimitri.scheftelowitsch (at) cs.tu-dortmund.de.
  • Dr. Bernd Zey 
    • Phone:  (+49)231 755-7735  
    • E-Mail: bernd.zey (at) tu-dortmund.de
  • Dr. Fritz Bökler
  • Andreas Bremer, M. Sc. psych.
  • Dr.-Ing. Iryna Dohndorf
  • Dr. Anna Ilyina
  • Dr. Jannis Kurtz
  • Dr. Thorsten Plewan
  • Dr. Johanna Renker
  • Dr.-Ing. Christian Schoppmeyer
  • Jessica Schwarz, Dipl.-Psych. 
    •  Phone: (+49)228 9435-491 
    • E-Mail: jessica.schwarz (at) fkie.fraunhofer.de
  • Thomas Siwczyk, Dipl.-Inf.
  • Dr. Nadine Wollenberg 
To top of page