DESIGN OF LINEAR MULTIVARIABLE SYSTEMS FOR STABILITY UNDER LARGE PARAMETER UNCERTAINTY.

The Sequential Return Difference method for the stabilization of linear multivariable systems is first extended for the case of perfectly known plant. Lower triangular controllers are introduced which prevent the need for constraints on the maximum allowed control effect. This approach is then extended to cope with large plant-parameter uncertainty where a systematic sequential-design technique is presented which quarantees the system stability for the whole uncertainty range. Under certain conditions, this technique also provides a means by which it is possible to convert a nonminimum-phase problem to a minimum-phase one, thus allowing the application of controllers with large gain that reduce the sensitivity of the system transfer to parameter uncertainty. A synthesis technique is then introduced which is based on the sequential-design principle; this quarantees stability under uncertainty and meets pre-specified requirements on the sensitivity of the system performance at low frequencies.