Constructing tight LFT uncertainty models for robust controller design

Many robust design techniques use LFT-models for uncertainty representation. On the other hand a wide class of practical control problems are formulated in terms of a family of linear state space models that capture the dynamic response of a nonlinear or time varying plant over a number of operating conditions. The way in which such LFT-models are constructed from a given family of state space models is very important, due to its direct impact on the conservatism of the overall robust design. Unfortunately this problem has received little attention in the robust control literature. In this paper a computational procedure based on singular value decomposition (SVD) and genetic algorithms (GA) is proposed. An example that illustrates the efficiency of the proposed technique is also included.