Optimal design of experiments for control: a preposterior viewpoint

This paper considers using experimental design in model identification to increase the predicted probability of closed-loop performance. The methodology assumes a Bayesian modeling viewpoint, where experimental input-output data is used off-line to characterize the probability distribution of the model parameters. Our approach to design of experiments is to select experimental inputs that “shape” a preposterior distribution of the model parameters such that a certain region in the model parameter space containing a pre-specified percentage of the preposterior density, denoted as an HPD region, is a subset of the region for closed-loop performance. Roughly speaking, the resulting experiments reduce model parameter variance in directions orthogonal to the performance set boundary. A missile autopilot example is used to illustrate the results