There are always discrepancies between design models and the actual physical systems or phenomena that they model. These discrepancies can be due to several factors including the approximation of complex models by simpler ones, the presence of unavoidable experimental errors when collecting data, or due to unknown or unmodeled effects. Regardless of their source, the perturbations can degrade the performance of otherwise optimal designs.
This possibility has motivated over the years numerous investigations into robust design techniques that are less sensitive to data perturbations such as total-least-squares and Hoo techniques. This talk discusses alternative design strategies for models with perturbations, including state-space models with uncertain dynamics. In comparison to TLS and Hoo methods, the resulting design procedures perform data regularization as opposed to data de-regularization; a property that avoids the need for continuous testing of existence conditions and is useful for applications involving online or real-time operation.
An application in the context of mobile wearable computing is described. Wearable computers are devices that are supposed to provide users with real-time access to information databases in a natural and unobtrusive manner. A critical factor for the success of wearable computers is their ability to quickly and reliably process sensory data. This talk illustrates the application of the regularized robust filtering theory to hand-gesture interface for wearable computing.