Control-Oriented Design Using Surrogate-Based Optimization and Existence Conditions for Robust Performance

The issue of cost function is actually quite critical to the inclusion of control synthesis for design optimization. Systems are evaluated on their ability to perform missions. Consequently, the design space must include all variables that affect such ability. The closed-loop operation of such systems suggests a decomposition of the design space into separate subspaces relating the plant dynamics and the controller. Surrogate models, also known as meta-models, are developed as approximations of parametrized models across a design space. These models provide information that is suitable for design optimization, but they incur a computational expense that is dramatically less than would be required to evaluate the high-fidelity model. The use of surrogate models is evaluated to optimize the design with low computational cost. Simultaneous design of an open-loop plant and controller is exceedingly challenging.