Abstract
The problem is considered of bounding the H ∞-norm of the transfer function T y1u1 relating the output y 1 to the input u 1 for the feedback configuration shown in Fig. 1, subject to the closed-loop system being internally stable. A new criterion for internal stability is derived in terms of the Q-parameterization, where Q=K(I + P22 K)-1. By means of such a stability theory, it is shown that the H ∞ control problem can be decomposed into two one-sided model-matching problems of the form considered in Part 1 (Hung 1989) of the paper. A closed-form state-space solution with a tight McMillan degree bound is obtained for H ∞ optimal or suboptimal controllers. An iterative algorithm for obtaining the H ∞ optimal controller is provided. Computationally, the algorithm requires the solution of only two algebraic Riccati equations in each iteration for which the upper bound on ∣∣T y1u1∣∣∞ is adjusted.