Vícekriteriální optimalizace s omezeními pomocí analýzy potenciálních ploch
Real-world optimization problems often involve both multiple conflicting objectives and constraints. Such constrained multiobjective problems (CMOPs) are generally hard to solve. While population-based metaheuristics, such as multiobjective evolutionary algorithms (MOEAs), are a successful approach for solving multiobjective optimization problems, the increasing number of objectives and the presence of constraints critically reduce their effectiveness. A powerful means for characterizing the optimization problems, tuning the algorithms and improving their performance is problem landscape analysis. In this project we will exend the concept of problem landscapes to multiobjective optimization with constraints and enhance the performance of evolutionary metaheuristics for continuous CMOPs. This will be achieved with new methods for problem landscape modeling, original approaches for identification and extraction of landscape features, a new test suite of CMOPs reflecting the properties of real-world problems and rigorous evaluation of the developed concepts.
Grantová agentura ČR
Mezinárodní grantové projekty Lead Agency (GF)
Information system of research, development and innovation (in Czech)