The goal of this article is to provide first-order necessary conditions of optimality for every local optimal pair (x~*, u~*) of the problem Minimize L(x, u), on all (x, u) ∈X*U subject to Ax-Bu f, where X, U, Y are Banach spaces, f ∈ Y, A: D(A) is contained in X → 2~Y, B:D(B) is contained → 2~Y are linear closed operators, L : X * U → IR is locally Lipschitz continuous or of the form L(x,u) = g(x) + h(u) where g:X → IR is locally Lipschitz continuous and h : U → IR∪{∞} is proper and convex. More precisely, the optimality conditions for (x~*, u~*) are partial deriv_xL(x~*, u~*) + A~* p 0, partial deriv_yL(x~*,u~*) -B~*q 0, p - q ∈(R(A) - R(B))~⊥, for some p,q ∈Y~*, Here Partial derivL(x~*, u~*) = (partial deriv_xL(x~*, u~*), partial deriv_yL(x~*,u~*)). Under additional hypothese these necessary conditions become: partial deriv_xL(x~*,u~*) + A~* p 0, partial deriv_yL(x~*,u~*) - B~*p 0.
展开▼
机译:本文的目的是为问题的每个局部最优对(x〜*,u〜*)提供一阶必要的最优性条件在所有(x,u)∈X*上使L(x,u)最小化U受Ax-Bu f约束,其中X,U,Y是Banach空间,f∈Y,A:X中包含D(A)→2〜Y,B:D(B)中包含→2〜Y是线性封闭算子,L:X * U→IR局部为Lipschitz连续或形式为L(x,u)= g(x)+ h(u)其中g:X→IR局部为Lipschitz连续,h:U→ IR∪{∞}是适当的且凸的。更准确地说,(x〜*,u〜*)的最优条件是偏导数xL(x〜*,u〜*)+ A〜* p 0,偏导数yL(x〜*,u〜*)-B〜* q 0,p-q∈(R(A)-R(B))〜⊥,对于某些p,q∈Y〜*,这里Partial derivL(x〜*,u〜*)=(partial deriv_xL(x〜 *,u〜*),部分deriv_yL(x〜*,u〜*))。在另外的假设下,这些必要条件变为:部分deriv_xL(x〜*,u〜*)+ A〜* p 0,部分deriv_yL(x〜*,u〜*)-B〜* p 0。
展开▼