Max-plus methods for nonlinear control and estimation. (Q819904)

This outstanding book discusses new theoretical approaches to the study of the control and estimation of continuous-time/continuous-space nonlinear systems to be a challenging problem. The main purpose of author is the presentation of some entirely new classes of numerical methods for the solution of nonlinear control problems (more generally, Hamilton-Jacobi-Bellman partial differential equations). This book is the first to provide, within a unified framework, a self-containted comprehensive mathematical theory of the max-plus algebra and some basics of functional analysis over the max-plus algebra. Topics include some of the results for max-plus function spaces which would best be described as max-plus functional analysis. The author presents several classes of nonlinear control problems, and nonlinear robust \(H^{\infty}\) control and estimation problems. The mathematical analysis of the problem covers existence, uniqueness and time evolution of the solution. This is carried out in the framework of the functional analysis over the max-plus algebra. The Dynamic Programming Principle approaches to solving nonlinear control problems is presented. It takes an operator mapping the value function (optimal cost as a function of system state) at one time to the value function at a later (or earlier) time. In the continuous-time case, if one takes the limit in the DPP as the time-interval goes to zero, the Dynamic Programming Equation is obtained. Moreover, for continuous-space problems, this DPE takes the form of an HJB PDE. Here PDE is a nonlinear, first order PDE. Some new numerical techniques (including a max plus eigenvector approach and an approach that avoids the curse-of dimensionality) are proposed. The max-plus-based methods belong to an entirely new class of numerical methods for the solution of nonlinear control problems and their associated HJB PDEs. The author's results show (the precise proofs of all theorems and lemmas are given) that the max-plus-based approaches lead to the solution operators for nonlinear HJB problems which are linear over the max-plus algebra, and this linearity is exploited in the construction of algorithms. The volume is primarily addressed to applied mathematicians working in the field of the control of nonlinear systems through the implementation of recent numerical methods. However, the book will also be useful for scientists from the application areas, in particular, applied scientists from engineering and physics.