A new algorithm for computing QFT bounds

An important step in quantitative feedback theory (QFT) design is the translation of problem data into the so-called QFT bounds. Initially, QFT practitioners relied on manual manipulation of plant templates over Nichols charts to obtain bounds. This tedious process has since been replaced by more efficient numerical algorithms. However, use of such algorithms is susceptible to the “curse of dimensionally” since plant templates are almost always defined by a finite approximation. A plant with 4 uncertain parameters may have to be approximated by, say, a set of 10⁴ members (a grid of 10 for each parameter). As a result, in most problems the designer is forced to trade-off between choosing a coarse plant grid to minimize the computational burden vs. a fine grid to maintain robustness (in the sense that the computed bounds are “close” to those for the true plant). This paper introduces a new technique which, improves robustness without increasing the computational burden