A pointwise constrained version of the Liapunov convexity theorem for vectorial linear first-order control systems (Q272247)

From MaRDI portal
scientific article
Language Label Description Also known as
English
A pointwise constrained version of the Liapunov convexity theorem for vectorial linear first-order control systems
scientific article

    Statements

    A pointwise constrained version of the Liapunov convexity theorem for vectorial linear first-order control systems (English)
    0 references
    0 references
    0 references
    0 references
    20 April 2016
    0 references
    The authors generalize the version of the Lyapunov convexity theorem for vectorial control systems driven by linear ordinary differential equations of first-order, in any dimension, by including a pointwise state-constraint. Let \(d\in\mathbb{N}\) and \(u^0,u^1,\dots, u^m\) be integrable functions in \(L^1((a,b),\mathbb{R}^d)\) with \((a,b)\in \mathbb{R}^2\), \(a< b\). Denote by \(U^{bb}(t)= \{u^0(t), u^1(t),\dots, u^m(t)\}\), \(t\in [a,b]\), the moving control-vertices set and \(U^{co}(t)= coU^{bb}(t)\), \(t\in [a,b]\), where \(bb\) stands for bang-bang (or true) control and \(co\) is the convex hull. If \(A^k\), \(0\leq k\leq p-1\), \(p\in\mathbb{N}\setminus\{0\}\), are \(d\times d\)-integrable matrices with entries \(A^k_{ij}\in L^1((a,b),\mathbb{R})\), consider the general \(p\)-th order linear ordinary differential operator \[ {\mathcal L}^p x(t)= x^{(p)}(t)+ A^{p-1}(t) x^{(p-1)}(t)+\cdots+ A^1(t) x'(t)+ A^0(t) x(t),\;t\in [a,b]. \] Consider the solution-spaces to the associated convex and nonconvex vectorial differential inclusions: \[ \begin{aligned} {\mathcal X}^{co} &= \{x\in{\mathcal X}^{pd};\;u(t)={\mathcal L}^p x(t)\in U^{co}(t)\text{ a.e. }t\in [a,b]\},\\ {\mathcal X}^{bb} &= \{x\in{\mathcal X}^{pd};\;u(t)={\mathcal L}^p x(t)\in U^{bb}(t)\text{ a.e. }t\in [a,b]\},\end{aligned} \] where \({\mathcal X}^pd= W^{p,1}([a,b],\mathbb{R}^d)\) is the \(p\)-th order Sobolev space. -- Let \[ \begin{aligned} \overline{{\mathcal X}}_a &= \{x\in{\mathcal X}^{pd};\;x^{(k)}(a)=\overline x^{(k)}(a),\;0\leq k\leq p-1\},\\ \overline{{\mathcal X}}_b &= \{x\in{\mathcal X}^{pd};\;x^{(k)}(b)= \overline x^{(k)}(b),\;0\leq k\leq p-1\},\end{aligned} \] where \(\overline x\in{\mathcal X}^{co}\) is a relaxed solution \(\overline{{\mathcal X}}=\overline{{\mathcal X}}_a\cap \overline{{\mathcal X}}_b\), and \(\overline{{\mathcal X}}^{co}={\mathcal X}^{co}\cap\overline{{\mathcal X}}\), \(\overline{{\mathcal X}}^{bb}={\mathcal X}^{bb}\cap\overline{{\mathcal X}}\) are subspaces consisting of relaxed or bang-bang solutions to the differential inclusion which are \(p\)-tangent to \(\overline x\) on the boundary \(\{a,b\}\). Consider the set \[ \Omega= \{\omega\in W^{p,1}([a,b],\mathbb{R}^d);\;\omega(t)\neq 0\text{ a.e. }t\in [a,b]\}, \] the function \(\omega\in\Omega\) being a moving directional vector. If \((\mathbb{R}^d,\langle\cdot,\cdot\rangle)\), where \(\langle\cdot,\cdot\rangle\) is the inner product in \(\mathbb{R}^d\), consider \[ \overline{{\mathcal X}}_-= \{x\in\overline{{\mathcal X}};\langle x(t),\omega(t)\rangle\leq \langle \overline{x}(t),\omega(t)\rangle,\;t\in [a,b]\}, \] the spaces of functions \(x\) being \(p\)-tangent to \(\overline x\) at \(a\) and \(b\), and \(\omega\)-dominated by \(\overline x\). -- If \[ y(t)=\langle x(t), \omega(t)\rangle,\quad \overline y(t)=\langle \overline{x}(t), \omega(t)\rangle,\quad t\in [a,b], \] then \(y(t)\leq\overline y(t)\), \(t\in [a,b]\). Consider the relaxed and bang-bang \(\omega\)-dominated solution sets \[ \overline{{\mathcal X}}^{co}_-={\mathcal X}^{co}\cap\overline{{\mathcal X}}_-,\quad \overline{{\mathcal X}}^{bb}_-= {\mathcal X}^{bb}\cap\overline{{\mathcal X}}_-, \] such that for every \(\overline x\in{\mathcal X}^{co}\) one has \(\overline{{\mathcal X}}^{bb}\neq\emptyset\). The authors study the existence for \(p=1\) of \(\omega\)-extremal \(b\)-weak bang-bang solutions, considering the differential operator of first order \[ {\mathcal L}^p x(t)= x'(t),\qquad t\in [a,b], \] and \[ {\mathcal L}^p_\omega y(t)= y^{(p)}(t)+ \mu^{p-1}_\omega(t) y^{(p-1)}(t)+\cdots \mu^1_\omega(t) y'(t)+ \mu^0_\omega(t) y(t),\;t\in [a,b], \] where \[ \mu^k_\omega\in L^1((a,b),\mathbb{R}),\qquad 0\leq k\leq p-1. \] Let \[ A^k(t)^T \omega= \mu^k_\omega(t),\qquad\text{a.e. }t\in [a,b],\quad 0\leq k\leq p-1, \] and \[ {\mathcal L}^p_\omega\overline y(t)\in co\{\rho^0(t),\dots, \rho^m(t)\},\qquad t\in [a,b], \] with \[ \rho(t)=\langle u(t),\omega(t)\rangle,\quad \rho^j(t)=\langle u^j(t),\omega(t)\rangle,\quad 1\leq j\leq m. \] Define \[ \begin{aligned} \overline{{\mathcal X}}^{co}_{\omega b-w} &= \{x\in{\mathcal X}^{co}\cap \overline{{\mathcal X}}_a;\;\langle x^k(b)-\overline x^k(b), \omega(b)\rangle= 0,\;0\leq k\leq p-1\},\\ \overline{{\mathcal X}}^{bb}_{\omega b-w} &= {\mathcal X}^{bb}\cap \overline{{\mathcal X}}^{co}_{\omega b-w}.\end{aligned} \] The authors prove the existence of bang-bang \(\omega\)-extremal unconstrained \(b\)-weak solutions: if \(p=1\) then there exist \[ \begin{aligned}\overline x^{\min}_{\omega b-w}\in\overline{{\mathcal X}}^{bb}_{\omega b-w},\qquad & \overline X^{\max}_{\omega b-w}\in \overline{{\mathcal X}}^{bb}_{\omega b-w},\\ \overline y^{\min}_{\omega b-w}= \langle\overline x^{\min}_{\omega b-w},\omega\rangle,\qquad & \overline y^{\max}_{\omega b-w}= \langle\overline x^{\max}_{\omega b-w}, \omega\rangle\end{aligned} \] and for every \(x\in\overline{{\mathcal X}}^{co}_{\omega b-w}\) one has \[ \overline y^{\min}_{\omega b-w}(t)\leq y(t)\leq \overline y^{\max}_{\omega b-w}(t),\quad t\in [a,b]. \] Associate to every \(x\in\overline{{\mathcal X}}^{co}_-\) the open set \[ {\mathcal O}_x= \{t\in (a,b);\;\langle x(t),\omega(t)\rangle< \langle\overline x(t), \omega(t)\rangle\} \] and define \[ \begin{aligned} \overline{{\mathcal X}}^{bb}_{--} &= \{x\in \overline{{\mathcal X}}^{bb}_-;\;{\mathcal O}_x\neq\emptyset\},\\ \overline{{\mathcal X}}^{co}_{--} &= \{x\in \overline{{\mathcal X}}^{co}_-;\;({\mathcal O}\neq\emptyset)\wedge (\exists{\mathcal L}^p x(t)\in U^{bb}(t)\text{ a.e. }t\in[a,b]\setminus{\mathcal O}_x\}.\end{aligned} \] The authors prove a necessary and sufficient condition for the existence of bang-bang \(\omega\)-dominated \(b\)-strong solutions: if \(p= 1\) then \[ (\overline{{\mathcal X}}^{bb}_{--}\neq\emptyset)\Leftrightarrow (\overline{{\mathcal X}}^{co}_{--}\neq\emptyset). \] They prove an alternative sufficient condition for the existence of bang-bang \(\omega\)-dominated \(b\)-strong solutions, a uniform approximation result for any \(\overline x\in{\mathcal X}^{co}\) by an \(\omega\)-increasing sequence of bang-bang \(\omega\)-dominated \(b\)-strong solutions, and a final result about product-increasing sequences of bang-bang solutions.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    Lyapunov convexity theorem
    0 references
    nonconvex linear differential inclusions
    0 references
    pointwise constraints
    0 references
    bang-bang solutions
    0 references
    linear boundary value control problems
    0 references
    0 references