ABSTRACT
Feedback controllers are designed to achieve certain performance specifications in the presence of both plant and signal uncertainty. Typical controller design formulations consider quadratic cost functions on the errors primarily for mathematical convenience. However, practical situations may dictate other kinds of measures. In particular, the peak values of error signals are oftenmore appropriate for describing desired performance. This can be a consequence of uniform tracking problems, saturation constraints, rate limits, or simply a disturbance rejection problem. In addition, disturbance and noise are in general bounded and persistent because they continue to act on the system as long as the system is in operation. Such signals are better described by giving information about both the signals’ frequency content and time-domain bounds on peak values.
