A Convex Optimization Approach to the Design of Multiobjective Discrete Time Systems

Anirudh Oberoi

Abstract

One of the most important contributions of robust control theory has been the devel opment of a new framework for the design and analysis of feedback systems satisfying mixed time-frequency specifications. This framework is given by the Linear Matrix Inequality (LMI) approach where design and analysis problems are posed as convex optimization problems subject to affine matrix constraints. Most of the focus in this area has been on continuous-time systems design with very few results for discretetime systems. One of the main contributions of this work is the development and implementation of a MATLAB toolbox for discrete-time controller design using the LMI approach. Another important contribution is the development of a new linear matrix inequality for peak-to-peak gain minimization that allows the use of projec tion formulas for l1-design. In order to illustrate the advantages and effectiveness of the LMI framework to multiobjective design problems it was applied to design a noise-shaping feedback coder. This nonlinear circuit is an important component of (Sigma) - (Delta) modulators. This work shows that a robust control approach based on LMIs provides a rigorous framework for the systematic analysis and design of these coders in contrast to existing ad hoc methods used traditionally for such designs.