Mixed H2/H∞ control for infinite dimensional systems

Abstract

The class of infinite dimensional systems often occurs when dealing with distributed parameter models consisting of partial differential equations. Although forming a comprehensive description, they mainly become manageable by finite dimensional approximations which likely neglect important effects, but underlies a certain structure. In contrast to common techniques for controlling infinite dimensional systems, this work focuses on using robust control methods. Thus, the uncertainty structure that occurs due to the discretization shall be taken into account particularly. Additionally, optimal performance measures can be included into the design process. The mixed H2/H∞ control approach handles the inclusion of disturbances and inaccuracies while guaranteeing specified energy or magnitude bounds. In order to include various of these system requirements, multi-objective robust control techniques based on the linear matrix inequality framework are utilized. This offers great flexibility concerning the formulation of the control task and results in convex optimization problems which can be solved numerically efficient by semi-definite programming. A flexible robot arm structure serves as the major application example during this work. The model discretization leads to an LTI system of specified order with an uncertainty model which is obtained by considering the concrete approximation impact and frequency domain tests. A structural analysis of the system model relates the neglected dynamics to a robust characterization. For the objective selection, stability shall be ensured under all expected circumstances while the aspects of optimal H2 performance, passive behavior and optimal measurement output selection are included. The undesirable spillover effect is thoroughly investigated and thus avoided.