Evolutionary Multiobjective Optimization


Many real-world problems involve simultaneous optimization of several incommensurable and often competing objectives. Typically, there is a trade-off of cost versus quality, where quality itself is defined in terms of further sub-criteria like performance, power dissipation, size, extensibility, etc. in the case of, e.g, embedded hardware/software system. There is usually no single optimal solution for any of these applications, but rather a set of alternative solutions. These solutions are optimal in the wider sense that no other solutions in the search space are superior to them when all objectives are considered. They are known as Pareto-optimal solutions.

The goal of our research in the area of evolutionary multiobjective optimization is:

  • to achieve a better understanding of existing approaches to multiobjective optimization by performing an extensive comparison on several test problems;
  • to extend the theory of multi-objective optimization in the following areas: run-time analysis, convergence proofs, comparison of different search methods, performance indicators.
  • to improve existing and to develop new, alternative multicriteria optimization methods;
  • to develop methods for indicator-based multiobjective search in high-dimensional spaces;
  • to apply multiobjective optimization techniques to the computer aided design of complex, heterogeneous hardware/software systems;
  • to provide a comprehensive programming framework on the basis of which new applications involving multiple, conflicting objectives can be implemented with minimal effort.

A comprehensive library of multi-objective search methods and associated statistical comparison methods are available at the PISA home page.

Some relevant publications on our work since 2000: Publications on Evolutionary Multiobjective Optimization, Publications on Biological Applications.

Further Information can be found via the home page of the system optimization research group.