A problem of convex analysis arising in optimal control theory (Q1813399)

Let \((\Omega,{\mathcal T},\mu)\) be a positive finite measure space, \(X\) a separable Banach space and \(L^ 1(X)\) the Banach space of all \(\mu\)- integrable vector functions \(x=x(.): \Omega\to X\). Let \(p: \Omega\times X\to R\) be a function \(\mu\)-measurable with respect to \(\omega\), sublinear with respect to \(x\) and such that \(p_ \omega(x)=p(\omega,x)\leq c\| x\|\), for all \((\omega,x)\in\Omega\times X\). The main result of the paper is the following: If \(\| x'\|\geq\varepsilon_ 0>0\) for all \(x'\in\partial p_ \omega\) (the subdifferential of \(p_ \omega\)) and all \(\omega\in\Omega\), then the conjugate \(K^*\) of the cone \(K=\{x=x(.)\in L^ 1(X): p(\omega,x(\omega))\leq 0\qquad \mu\hbox{-a.e. on }\Omega\}\) is given by \(K^*=\{-\lambda y': \lambda\geq0, y'(\omega)=\alpha(\omega)\cdot x'(\omega)\), \(\alpha\in{\mathcal P}\), \(x'=x'(.)\in L^ \infty(X^*)\), \(x'(\omega)\in\partial p_ \omega\), \(\forall\omega\in\Omega\}\), where \({\mathcal P}=\{\alpha\in L^ \infty: \alpha\geq0,\quad \int_ \Omega\alpha(\omega)\cdot\mu(d\omega)=1\}\).