Searching for knee regions in multi-objective optimization using mobile reference points

摘要: Evolutionary algorithms have amply demonstrated their effectiveness and efficiency in approximating the Pareto front of different multi-objective optimization problems. Fewer attentions been paid to search for preferred parts according decision maker preferences. Knee regions are special portions containing solutions having maximum marginal rates return, i.e., which an improvement one objective implies a severe degradation at least another one. Such characteristic makes knee particular interest practical applications from perspective. In this paper, we propose new updating strategy reference points based evolutionary algorithm forces latter focus on regions. The proposed idea uses set mobile guiding towards extent obtained could be controlled by means user-defined parameter. verification approach is assessed two- three-objective knee-based test problems priori interactively. results promising.