Differential games and nonlinear \({\mathcal H}_\infty\) control in infinite dimensions (Q2706152)

From MaRDI portal





scientific article
Language Label Description Also known as
English
Differential games and nonlinear \({\mathcal H}_\infty\) control in infinite dimensions
scientific article

    Statements

    0 references
    0 references
    19 March 2001
    0 references
    viscosity solutions
    0 references
    nonlinear semigroups
    0 references
    accretive operators
    0 references
    dynamic programming
    0 references
    nonlinear \(\mathcal H_\infty\) control
    0 references
    optimality principles
    0 references
    nonlinear unbounded infinite-dimensional system
    0 references
    state constraints
    0 references
    family of strategies
    0 references
    differential game
    0 references
    Differential games and nonlinear \({\mathcal H}_\infty\) control in infinite dimensions (English)
    0 references
    This paper deals with an \(H_\infty\) control problem for a nonlinear unbounded infinite-dimensional system with state constraints. For Hilbert spaces \(U\) (=control set) and \(W\) (=disturbance set), the authors considered the following controlled system \(y(t)= y(t,x,u,w)\) in a Hilbert space \(H\); \(y'(t)+ Ay(t)\ni f(y(t))+ Bu(t)+ Cw(t)\) with \(y(0)= x(\in\overline{D(A)})\), where \(u\in L^2_{\text{loc}}(0,\infty, U)\), \(w\in L^2_{\text{loc}}(0,\infty, W)\) and \(A\) is a maximal monotone in \(H\). For fixed \(\gamma> 0\) and running cost function \(g\), \(\{\alpha_x, x\in \overline{D(A)}\}\) is called a family of strategies of the controller if \(\alpha_x\): \(L^2_{\text{loc}}(0,\infty, W)\to L^2_{\text{loc}}(0,\infty, U)\) is causal and the condition NEWLINE\[NEWLINE\int^T_0(g(y(t, x,\alpha_x[w], w))+\|\alpha_x[w](t)\|^2) dt\leq \gamma^2 \int^T_0\|w(t)\|^2 dt+ K(x),\;\forall T,w,NEWLINE\]NEWLINE holds for some nonnegative function \(K\) with \(K(0)= 0\). Defining the value function of the differential game by NEWLINE\[NEWLINEV_\gamma(x):= \inf_\alpha \sup_w \sup_T \int^T_0 (g(y(t, x,\alpha_x[w], w))+ \|\alpha_x[w](t)\|^2- \gamma^2\|w(t)\|^2) dt,NEWLINE\]NEWLINE the authors show that (1) the Hamilton-Jacobi-Isaacs (HJI) equation associated with the differential game has a viscosity supersolution if and only if there is a family of strategies, (2) the \(H_\infty\) problem can be solved if and only if the HJI equation has a positive definite viscosity supersolution vanishing and continuous at 0. They apply the results to the one phase Stefan problem.
    0 references
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references