Linear Lyapunov-Krotov functions and sufficient conditions for optimality in the form of the maximum principle (Q1413071): Difference between revisions

The main result of the paper is Theorem 1 containing a sufficient optimality condition for a given extremal \((\overline u(.),\overline x(.))\) (satisfying the well-known Pontryagin's Maximum Principle), of an optimal control problem which consists in the \textit{minimization} of the functional \[ J(u(.),x(.)):=l_0(b(x(.))), \;b(x(.)):=(t_0,x(t_0),t_1,x(t_1)) \] subject to: \[ x'(t)=f(t,x(t),u(t)), \;u(t)\in U, \;l(b(x(.)))\leq 0, \;k(b(x(.)))=0 \] defined by smooth data. To the given extremal one associates a family \(\{\psi^a(.); \;a\in {\mathcal A}\}\) of adjoint variables (``coextremals'' in authors' terminology) that satisfy a so-called ``extended maximum condition'' \[ \max_{x,u}[\langle \psi(t),f(t,x,u)\rangle +\langle \psi'(t),x\rangle ]=\langle \psi(t),f(t,\overline x(t), \overline u(t))\rangle +\langle \psi'(t),\overline x(t)\rangle \] and one proves (rather trivially) that if \(\overline b:=(\overline t_0, \overline x(\overline t_0),\overline t_1,\overline x(\overline t_1))\) is a solution of the finite-dimensional optimization problem of {minimizing} \(l_0(b)\) subject to: \[ l(b)\leq 0, \;k(b)=0, \;\langle \psi^a(t_1),x_1-\overline x(t_1)\rangle - \langle \psi^a(t_0),x_0-\overline x(t_0)\rangle \leq 0, \;b=(t_0,x_0,t_1,x_1), \;a\in {\mathcal A} \] then \((\overline u(.),\overline x(.))\) is optimal. Sufficient optimality conditions of a similar nature have been obtained in [\textit{V. I. Blagodatskih}, Sov. Math., Dokl. 17 (1976), 1680--1683 (1977); translation from Dokl. Akad. Nauk SSSR 231, 1041--1044 (1976; Zbl 0399.49011)], [\textit{Şt. Mirică}, An. Univ. Bucur., Mat. 32, 47--56 (1983; Zbl 0534.49019)] and in [\textit{R. T. Rockafellar}, Lect. Notes Math. 680, 219--257 (1978; Zbl 0385.49001)].