Computing Feedback Laws for Linear Systems with a Parallel Pieri Homotopies.pptVIP

Computing Feedback Laws for Linear Systems with a Parallel Pieri Homotopies Yusong Wang Department of Math, Stat CS University of Illinois at Chicago E-mail: ywang25@ HPSEC04, Montreal, Canada Aug. 15, 2004 Joint work with Jan Verschelde Outline of the talk The extension of PHCpack with a parallel path tracker. ? Introduction of Homotopy ? The reason we need a parallel path tracker Parallel Pieri Homotopies ? Relation between Parallel Pieri Homotopies and output feedback pole placement problem ? Difficulty compared with general homotopy algorithm ? Localization pattern of the solutions ? Definition of Pieri Homotopies ? Root-counting with poset and tree structure ? Parallel Pieri algorithm and its applications. Introduction 1 Numerical Homotopy Algorithm If we wish to solve f(x) = 0, then we construct another polynomial g(x) = 0 whose solutions are known. Consider the homotopy H(x,t) = r(1 - t)g(x) + t f(x) = 0, r ∈ ?. By continuation, we trace the paths at the known solutions of g(x) = 0 to the desired solutions of f(x) = 0, for t from 0 to 1. T.Y. Li. Numerical solution of polynomial systems by homotopy continuation methods. Handbook of Numerical Analysis. Volume XI. pages 209--304, 2003. 2 Why do we need a parallel path tracker? A) For large problems, the computational workload increases dramatically. Difficulty of Pieri Homotopies Tracking all paths defined by one homotopy is “embarassingly parallel”, as the tasks no longer communicate with each other once they are created. I. Foster. Designing and Building Parallel Programs. Addison-Wesley Publishing 1995. In Pieri homotopies, we have a hierarchy of homotopies, which is described by a poset. The solution at the end of one path may serve as the start solution for another path. So the tasks communicates in a predictable pattern, which is no longer embarassingly parallel. Pieri Homotopies and control problem The