The autonomous linear quadratic control problem. Theory and numerical solution (Q1189408)

This book presents a survey of the state of the art in the theory and numerical treatment of general linear quadratic optimal control problems with differential algebraic equation constraints. Special emphasis is devoted to the descriptor case which has attracted much interested in the last ten years. The continuous time problem is considered as well as the discrete time problem. In the theoretical part of the book the application of the maximum principle is discussed, including analysis of the resulting boundary value problems. Optimal stabilizing feedback solutions are obtained via the solution of algebraic Riccati equations or Riccati differential equations. Some of the results on Riccati equations are extended to the general descriptor case. A number of canonical forms like Schur forms, Hessenberg forms and triangular decompositions are discussed which simplify the construction of numerical methods. The numerical part of the book starts with a general defect correction method due to the author. It can be combined with other methods as a procedure for iterative refinement. As a particular example Newton's method is discussed in detail. Based on the idea of deflating subspaces, the sign function method is presented. Unitary transformations to transform the matrix pencils to a Schur like form are described. Algorithms for special Hamiltonian or symplectic eigenvalue problems are reported, using the symplectic geometry, which is hidden in these problems. A combination algorithm for real Hamiltonian and symplectic matrices is due to the author himself. All presented methods are illustrated by means of a decision flow chart which leads the reader like an expert system. The book will help to increase the cooperation between applied mathematicians and engineers in the field of optimal control.