Vector-valued versus scalar-valued figures of merit in \(H^{\infty}\)- feedback system design (Q1119523)

In the theory of design of feedback systems for a linear time-invariant plant P(s), the engineer is confronted with competing design objectives to choose a stabilizing compensator C so as to make the \(H^{\infty}\)- norm of both the sensitivity function \(S=(I+PC)^{-1}\) and the complementary sensitivity function \(T=I-S=PC(I+PC)^{-1}\) small. A standard compromise procedure is to choose weighting functions \(W_ 1\) and \(W_ 2\) and then design C so as to minimize a weighted combination \((\| W_ 1S\|^ P+\| W_ 2T\|^ P)^{1/P}\). The authors propose that what is really needed is a design which makes the pair \((\| W_ 1S(C_ 0)\|,\| W_ 2T(C_ 0)\|)\) a minimal element in a partial ordering on the planar convex set of achievable pairs \((\| W_ 1S\|,\| W_ 2T\|)\) and gives advice on how to choose the weights so that the above scalar criterion directs the design to achieving this vectorial performance objective.