Design by continuous collaboration between manual and automatic optimization

Numerical optimization is traditionally viewed as a machine centric activity. This view dominates the majority of numerical optimization packages today, where user interaction is normally limited to the problem definition phase or visualization of the results with little or no interaction at design or run time. Surprisingly we are surrounded by many examples of successful engineering systems which allow human interaction at run time, e.g. automobiles, aircraft etc. In fact, Integrated Human Machine Systems (IHMS) and dedicated engineering design groups have already shown that the distribution of the intelligent function between the human and artificial agents at design time leads to a more effective utilization of their complementary capabilities. This paper discusses the implementation of a generic semi-automatic optimization concept in which the human designer continuously collaborates with a numerical agent to navigate the design space and modify it when necessary. The concept allows human interaction at various levels of automation. The potential of this approach is shown by way of three human-in-the-loop optimization examples: . conceptual design optimization of subsonic aircraft; . optimization of trajectory of a Mars rover vehicle; . configuration optimization of a multistage rocket. This technique is located in a multidisciplinary area formed by the cross section of Visualization, Numerical Optimization and Computational Steering.