Adaptation in the presence of unstructured uncertainty

This paper examines the H^∞/l²-induced norm robust performance problems for sets of time-invariant plant uncertainty generated by a combination of an priori unstructured uncertainty assumptions and a posteriori identification data. Two cases are considered: time-invariant feedback for fixed sets of such plant uncertainty; and time-varying feedback for sets of plant uncertainty which contract with evolving data. In the latter case it is shown, under certain conditions, that feedback can be synthesised in such a way as to satisfy a formal notion of adaptation