On \(\varepsilon\)-optimal controls for state constraint problems (Q1591443)

From MaRDI portal
scientific article
Language Label Description Also known as
English
On \(\varepsilon\)-optimal controls for state constraint problems
scientific article

    Statements

    On \(\varepsilon\)-optimal controls for state constraint problems (English)
    0 references
    0 references
    0 references
    1 October 2001
    0 references
    As the authors state in the Introduction, the aim of this paper is to extend to state constrained problems the method in [\textit{F. H. Clarke, Yu. S. Ledyaev, E. D. Sontag} and \textit{A. I. Subbotin}, IEEE Trans. Aut. Contr. 42, No. 10, 1394-1407 (1997; Zbl 0892.93053)] to prove the existence of \textit{sub-optimal feedback controls} for problems of the form: \[ u(x):=\displaystyle \inf_{\alpha(.)}J(x,\alpha(.)), \;J(x,\alpha(.)):= \int_0^\infty e^{-t} f(X(t;x,\alpha(.)),\alpha(t)) dt, \;x\in \overline{\Omega} \subset R^N, \] subject to: \[ X'(t)=g(X(t),\alpha(t)), \;X(0)=x, \quad X(t):=X(t;x,\alpha(.))\in \overline{ \Omega} \;\forall \;t\geq 0 \] where \(\Omega \subset R^N\) is a bounded open subset, \(\overline{\Omega}:= \text{Cl}(\Omega) \) and \(\alpha (.):[0,\infty)\to A\subset R^m\) are measurable controls. The main result of the paper is Theorem 5.1 stating, essentially, that under some (rather restrictive) hypotheses on the boundary \(\partial \Omega\), for each \(\varepsilon >0\) there exist \(\widehat {\tau}>0\) and a mapping \(\widehat{\alpha}_\varepsilon (.):\overline{\Omega}\to A\) that is an \(\varepsilon\) -suboptimal feedback control in the following sense: for any \(x\in \overline{\Omega}\) the ``piecewise constant'' mapping \(\alpha_{\varepsilon ,x}(.)\) defined by: \[ \alpha_{\varepsilon ,x}(t):= \widehat{\alpha}_\varepsilon (x_k) \text{ if }t\in [k\widehat{\tau},(k+1) \widehat{\tau}), \;x_0=x, \;x_{k+1}=X(\widehat {\tau};x_k,\widehat{\alpha}_\varepsilon(x_k)), \;k=0,1,2,\dots \] is an \(\varepsilon\)-optimal control (with respect to the initial point \(x\)) in the sense that: \[ u(x)\leq J(x,\alpha_{\varepsilon ,x}(.))<u(x)+\varepsilon, \] where \(u(.)\) is the value function of the problem. The proof of the main result takes some 4 pages and is based on a large number of very technical auxiliary results. Related results, in the slightly different context of ``proximal analysis'' may be found in [\textit{F. Clarke, Yu. S. Ledyaev} and \textit{A. I. Subbotin}, Tr. Mat. Inst. Steklova 224, 165-186 (1999; Zbl 0965.49022)].
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    optimal control
    0 references
    infinite horizon
    0 references
    sub-optimal feedback control
    0 references
    value function
    0 references
    Hamilton-Jacobi-Bellman equation
    0 references
    viscosity solution
    0 references