From Robust Design to Evolvable Systems Engineering – University of Copenhagen

From Robust Design to Evolvable Systems Engineering

Abstract:

Robustness is a system's property that has received increasing attention in the recent years both in engineering as well as in the life sciences. In the presentation, different measures and
approaches to robustness will be discussed in the context of applying natural computation to engineering systems. Theoretical results for evolution strategies show that it is more efficient to use larger population sizes than to sample for robust optimization. At the same time, including robustness in the system design process can lead to topological changes of the quality landscape that can pose problems to optimization methods that avoid resampling.

In the second part of the presentation, principles of Systems Engineering will be discussed and extended to include elements of temporal and spatial variability. The term evolvability in evolutionary biology is used to describe the system's ability to constructively react to environmental changes and to continuously improve. Evolvability in the context of systems engineering will be discussed to reconcile the holistic view with the practical need of problem decomposition.

Examples from aerodynamic design optimization will be used to emphasize the practical aspects of robust and evolvable optimization.

About Bernhard Sendhoff

  • Diploma Physics: Ruhr-Universität Bochum, Germany and University of Sussex, UK
  • PhD: Institut für Neuroinformatik, Ruhr Universität Bochum, Germany
  • Honorary Professor, University of Birmingham, UK
  • Honorary Professor, Technical University of Darmstadt, Germany
  • President, Honda Research Institute Europe GmbH