Nikola Aulig, "Evolutionary Optimization of the Robustness and Evolvability of Design Solutions", 2011.

In the case of an optimization subject to several environmental conditions, conventional optimizations can either provide several specialized solutions or a single robust solution with usually lower peak performance. This thesis proposes “evolvability” as an alternative optimization objective, in order to generate design solutions, which can be adapted to different conditions through small and cheap changes. Concepts of evolvability in evolutionary biology and evolutionary optimization are reviewed, and two evolvability measures are deduced in this work. Those are implemented and evaluated on a spline target fit shape optimization. Hereby, the optimization for evolvability identifies parameters, which have to be adapted and reveals structural differences between robust and evolvable solutions. Motivation for the thesis is the practical problem of adapting an underbody design to several car models, in order to minimize the drag of the overall vehicles. In the practical part therefore, the developed concept is applied to an underbody optimization and compared to conventional approaches. In doing so, the findings of the shape optimization could be confirmed and therewith the feasibility of the optimization for evolvability for a realistic optimization problem.