On time-minimal distributed control of vibrations (Q1117880)

From MaRDI portal
scientific article
Language Label Description Also known as
English
On time-minimal distributed control of vibrations
scientific article

    Statements

    On time-minimal distributed control of vibrations (English)
    0 references
    0 references
    1989
    0 references
    The author considers a control system described by an abstract wave equation of the form \(y''(t)+Ay(t)=f(t)\), where A is a self adjoint positive definite linear operator in a Hilbert space H. Let \(M>0\). The time optimal (null controllability) problem, referred to below as Problem 1, is this: given \((y_ 0,y_ 1)\in D(A^{1/2})\times H\), find among all the admissible controls, that is, controls that satisfy \(\| f(t)\| \leq M\), the (optimal) control \(\bar f_ M(t)\) that transfers the system from the initial state \((y_ 0,y_ 1)\) to the final equilibrium state (0,0) in minimum time T(M). This paper is an important contribution to the problem of numerical computation of the optimal control \(\bar f_ M(t)\). Its central idea is to consider another optimal control problem (Problem 2): given \(T>0\), find the (optimal) control \(\tilde f_ T(t)\) that transfers the system from the initial state \((y_ 0,y_ 1)\) to the final equilibrium state (0,0) minimizing \[ M(f)= \sup_{0\leq t\leq T}\| f(\cdot)\|. \] The author shows the following relation between the two problems: if \(\tilde f_ T(t)\) is the solution of Problem 2 for \(T=T(M)\), then \(\tilde f_ T(t)=\bar f_ M(t)\) a.e. (in particular M(\~f\({}_ T)=M\), and \(\tilde f_ T\) is the only control with this property). Accordingly, the calculation of \(\bar f_ M(t)\) can be carried out as follows: (a) compute \(\tilde f_ T\) for different values of T; (b) compute T such that \(\| \tilde f_ T\| =M\). Step (a) is an unconstrained minimization problem and can be implemented numerically; step (b) is a root finding problem that can be handled by any method that does not require derivatives (say secant or Müller). Other methods of computing optimal controls for Problem 1 exist (for instance, penalization methods). However, their numerical implementation requires solving nonlinear constrained minimization problems which may not behave properly as discretization steps tend to zero. To the reviewer's knowledge, the present method is the only one that has proved to be numerically successful for the problem at hand.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    self adjoint positive definite linear operator
    0 references
    Hilbert space
    0 references
    null controllability
    0 references
    numerical computation
    0 references
    optimal control
    0 references
    0 references