The development of many-objective evolutionary algorithms has facilitated solving complex design optimization problems, that is, optimization problems with four or more competing objectives. The outcome of many-objective optimization is often a rich set of solutions, including the non-dominated solutions, with varying degrees of tradeoff amongst the objectives, herein referred to as the trade space. As the number of objectives increases, exploring the trade space and identifying acceptable solutions becomes less straightforward. Visual analytic techniques that transform a high-dimensional trade space into two-dimensional (2D) presentations have been developed to overcome the cognitive challenges associated with exploring high-dimensional trade spaces. Existing visual analytic techniques either identify acceptable solutions using algorithms that do not allow preferences to be formed and applied iteratively, or they rely on exhaustive sets of 2D representations to identify tradeoffs from which acceptable solutions are selected. In this paper, an index is introduced to quantify tradeoffs between any two objectives and integrated into a visual analytic technique. The tradeoff index enables efficient trade space exploration by quickly pinpointing those objectives that have tradeoffs for further exploration, thus reducing the number of 2D representations that must be generated and interpreted while allowing preferences to be formed and applied when selecting a solution. Furthermore, the proposed index is scalable to any number of objectives. Finally, to illustrate the utility of the proposed tradeoff index, a visual analytic technique that is based on this index is applied to a Pareto approximate solution set from a design optimization problem with ten objectives.

This content is only available via PDF.
You do not currently have access to this content.