Abstract
In this paper, by using several variational techniques and a dual gap-type functional, we study weak sharp solutions associated with a controlled variational inequality governed by convex path-independent curvilinear integral functional. Also, under some hypotheses, we establish an equivalence between the minimum principle sufficiency property and weak sharpness for a solution set of the considered controlled variational inequality.
Supported by University Politehnica of Bucharest, Bucharest, Romania (Grant No. MA51-18-01).
1 Introduction
Convexity theory is an important foundation for studying a wide class of unrelated problems in a unified and general framework. Based on the notion of unique sharp minimizer, introduced by Polyak [11], and taking into account the works of Burke and Ferris [2], Patriksson [10], following Marcotte and Zhu [8], the variational inequalities have been strongly investigated by using the concept of weak sharp solution. We mention, in this respect, the works of Wu and Wu [15], Oveisiha and Zafarani [9], Alshahrani et al. [1], Liu and Wu [6] and Zhu [16].
In this paper, motivated and inspired by the ongoing research in this area, we introduce and investigate a new class of scalar variational inequalities. More precisely, by using several variational techniques presented in Clarke [3], Treanţă [12, 13] and Treanţă and Arana-Jiménez [14], we develop a new mathematical framework on controlled continuous-time variational inequalities governed by convex path-independent curvilinear integral functionals and, under some conditions and using a dual gap-type functional, we provide some characterization results for the associated solution set. As it is very well-known, the functionals of mechanical work type, due to their physical meaning, become very important in applications. Thus, the importance of this paper is supported both from theoretical and practical reasonings. As well, the ideas and techniques of this paper may stimulate further research in this dynamic field.
2 Notations, Working Hypotheses and Problem Formulation
In this paper, we will consider the following notations and working hypotheses:
\( \displaystyle \blacktriangleright \) two finite dimensional Euclidean spaces, \( \displaystyle \mathbb {R}^{n} \) and \( \displaystyle \mathbb {R}^{k} \);
\( \displaystyle \blacktriangleright \) \( \displaystyle \varTheta \subset \mathbb {R}^{m} \) is a compact domain in \( \displaystyle \mathbb {R}^{m} \) and the point \( \displaystyle \varTheta \ni t = (t^{\beta }), \, \beta = \overline{1,m} \), is a multi-parameter of evolution;
\( \displaystyle \blacktriangleright \) for \( \displaystyle t_{1} = (t_{1}^{1},\ldots ,t_{1}^{m}), \; t_{2} = (t_{2}^{1},\ldots ,t_{2}^{m}) \) two different points in \( \displaystyle \varTheta \), let \( \displaystyle \varTheta \supset \varUpsilon : t = t(\theta ), \theta \in [a,b] \) (or \( \displaystyle t \in \overline{t_{1},t_{2}} \)) be a piecewise smooth curve joining the points \( \displaystyle t_{1} \) and \( \displaystyle t_{2} \) in \( \displaystyle \varTheta \);
\( \displaystyle \blacktriangleright \) for \( \displaystyle \mathcal {U} \subseteq \mathbb {R}^{k}, \; \mathcal {P}:= \varTheta \times \mathbb {R}^{n} \times \mathcal {U} \) and \( \displaystyle i = \overline{1,n}, \; \beta = \overline{1,m}, \; \varsigma = \overline{1,q}\), we define the following continuously differentiable functions
\( \displaystyle \blacktriangleright \) for \( \displaystyle x_{\beta } := \frac{\partial x}{\partial t^{\beta }}, \; \beta = \overline{1,m} \), let \( \displaystyle \overline{\mathcal {\mathbf {X}}} \) be the space of piecewise smooth state functions \( \displaystyle x: \varTheta \rightarrow \mathbb {R}^{n} \) with the norm
\( \displaystyle \blacktriangleright \) also, denote by \( \displaystyle \overline{\mathcal {\mathbf {U}}} \) the space of piecewise continuous control functions \( \displaystyle u: \varTheta \rightarrow \mathcal {U} \), endowed with the uniform norm \( \displaystyle \parallel \cdot \parallel _{\infty } \);
\( \displaystyle \blacktriangleright \) consider \( \displaystyle \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \) equipped with the Euclidean inner product
and the induced norm;
\( \displaystyle \blacktriangleright \) denote by \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) a nonempty, closed and convex subset of \( \displaystyle \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \), defined as
where \( \displaystyle x, u \) are the simplified notations for \( \displaystyle x(t), u(t) \) and \(x_{1}\) and \(x_{2}\) are given;
\( \displaystyle \blacktriangleright \) assume the continuously differentiable functions \( \displaystyle V_{\beta } = \left( V^{i}_{\beta }\right) , \; i = \overline{1,n}, \; \beta = \overline{1,m}, \) satisfy the following conditions of complete integrability
where \( \displaystyle D_{\zeta } \) denotes the total derivative operator;
\( \displaystyle \blacktriangleright \) for any two q-tuples \( \displaystyle a = \left( a_{1}, ..., a_{q} \right) , b = \left( b_{1}, ..., b_{q} \right) \) in \( \displaystyle \mathbb {R}^{q} \), the following rules
are assumed.
Note. Further, in this paper, it is assumed summation on repeated indices.
In the following, \( \displaystyle J^{1}(\mathbb {R}^{m}, \mathbb {R}^{n}) \) denotes the first-order jet bundle associated to \( \displaystyle \mathbb {R}^{m} \) and \( \displaystyle \mathbb {R}^{n} \). For \( \displaystyle \beta = \overline{1,m} \), we consider the real-valued continuously differentiable functions (closed Lagrange 1-form densities) \( \displaystyle l_{\beta }, s_{\beta }, r_{\beta }: J^{1}(\mathbb {R}^{m}, \mathbb {R}^{n}) \times \mathcal {U} \rightarrow \mathbb {R} \) and, for \( \displaystyle (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \), define the following path-independent curvilinear integral functionals:
Definition 2.1
The scalar functional \( \displaystyle L(x,u) \) is called convex on \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) if, for any \( \displaystyle (x,u), (x^{0},u^{0}) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \), the following inequality
is satisfied.
Definition 2.2
For \( \displaystyle \beta = \overline{1,m} \), the variational (functional) derivative \( \displaystyle \delta _{\beta } L(x,u) \) of the scalar functional \( \displaystyle L: \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \rightarrow \mathbb {R}, \; L(x,u) = \int _{\varUpsilon } l_{\beta } \left( t, x, x_{\vartheta }, u \right) dt^{\beta } \), is defined as
with
and, for \( \displaystyle (\psi , \varPsi ) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \), with \( \displaystyle \psi (t_{1}) = \psi (t_{2}) = 0 \), the following relation
is satisfied.
Working assumptions. (i) In this work, it is assumed that the inner product between the variational derivative of a scalar functional and an element \( \displaystyle (\psi , \varPsi ) \) in \( \displaystyle \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \) is accompanied by the condition \( \displaystyle \psi (t_{1}) = \psi (t_{2}) = 0 \).
(ii) Assume that
is an exact total differential and satisfies \( \displaystyle U(t_{1}) = U(t_{2}) \).
At this point, we have the necessary mathematical tools to formulate the following controlled variational inequality problem: find \( \displaystyle (x^{0},u^{0}) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) such that
for any \( \displaystyle (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \). The dual controlled variational inequality problem associated to (CVIP) is formulated as follows: find \( \displaystyle (x^{0},u^{0}) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) such that
for any \( \displaystyle (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \).
Denote by \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) and \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})_{*} \) the solution set associated to (CVIP) and (DCVIP), respectively, and assume they are nonempty.
Remark 2.1
As it can be easily seen (see (ii) in Working assumptions), we can reformulate the above controlled variational inequality problems as follows: find \( \displaystyle (x^{0},u^{0}) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) such that
respectively: find \( \displaystyle (x^{0},u^{0}) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) such that
In the following, in order to describe the solution set \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) associated to (CVIP), we introduce the following gap-type path-independent curvilinear integral functionals.
Definition 2.3
For \( \displaystyle (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}}\), the primal gap-type path-independent curvilinear integral functional associated to (CVIP) is defined as
and the dual gap-type path-independent curvilinear integral functional associated to (CVIP) is defined as follows
For \( \displaystyle (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \), we introduce the following notations:
In the following, in accordance with Marcotte and Zhu [8], we introduce some central definitions.
Definition 2.4
The polar set \( \displaystyle (\mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}})^{\circ } \) associated to \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) is defined as
Definition 2.5
The normal cone to \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) at \( \displaystyle (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \) is defined as
and the tangent cone to \( \displaystyle \mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}} \) at \( \displaystyle (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \) is \( \displaystyle T_{\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}}}(x,u) = \left[ N_{\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}}}(x,u)\right] ^{\circ }\).
Remark 2.2
By using the definition of normal cone at \(\displaystyle (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \), we observe the following: \( \displaystyle (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \Longleftrightarrow \left( -\frac{\delta _{\beta } L}{\delta x^{*}}, -\frac{\delta _{\beta } L}{\delta u^{*}}\right) \in N_{\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}}}(x^{*},u^{*})\).
3 Preliminary Results
In this section, in order to formulate and prove the main results of the paper, several auxiliary propositions are established.
Proposition 3.1
Let the path-independent curvilinear integral functional \( \displaystyle L(x,u) \) be convex on \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \). Then:
(i) the following equality
is fulfilled, for any \( \displaystyle (x^{1}, u^{1}), (x^{2}, u^{2}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \);
(ii) \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \subset (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})_{*} \).
Remark 3.1
The property of continuity for the variational derivative \( \displaystyle \delta _{\beta } L(x,u) \) implies \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})_{*} \subset (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}\). By Proposition 3.1, we conclude \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} = (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})_{*} \). Also, the solution set \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})_{*} \) associated to (DCVIP) is a convex set and, consequently, the solution set \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) associated to (CVIP) is a convex set.
Proposition 3.2
Let the path-independent curvilinear integral functional \( \displaystyle R(x,u) \) be differentiable on \( \displaystyle \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \). Then the following ineguality
is satisfied, for any \( \displaystyle (x,u), (v,\mu ) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}}, \;(y,w) \in \mathcal {Z}(x,u) \).
Proposition 3.3
Let the path-independent curvilinear integral functional \( \displaystyle R(x,u) \) be differentiable on \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) and the path-independent curvilinear integral functional \( \displaystyle L(x,u) \) be convex on \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \). Also, assume the following implication
is true, for any \( \displaystyle (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}, \; (v,\mu ) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}}, \;(z,\nu ) \in \mathcal {Z}(x^{*},u^{*}) \). Then \( \displaystyle \mathcal {Z}(x^{*},u^{*}) = (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}, \; \forall (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \).
4 Main Results
In this section, taking into account the preliminary results established in the previous section, we investigate weak sharp solutions for the considered controlled variational inequality governed by convex path-independent curvilinear integral functional. Concretely, following Marcotte and Zhu [8], in accordance with Ferris and Mangasarian [4], the weak sharpness property of \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) associated to (CVIP) is studied. In this regard, two characterization results are established.
Definition 4.1
The solution set \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) associated to (CVIP) is called weakly sharp if there exists \( \displaystyle \gamma > 0 \) such that
(see \( \displaystyle int (Q) \) the interior of the set Q and B the open unit ball in \( \displaystyle \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \)), or, equivalently,
for all \(\displaystyle (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}\).
Lemma 4.1
There exists \( \displaystyle \gamma > 0 \) such that
if and only if
The first characterization result of weak sharpness for \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is formulated in the following theorem.
Theorem 4.1
Let the path-independent curvilinear integral functional \( \displaystyle R(x,u) \) be differentiable on \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) and the path-independent curvilinear integral functional \( \displaystyle L(x,u) \) be convex on \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \). Also, assume that:
(a) the following implication
is true, for any \( \displaystyle (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}, \; (v,\mu ) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}}, \; (z,\nu ) \in \mathcal {Z}(x^{*},u^{*}) \);
(b) \( \displaystyle \left( \frac{\delta _{\beta } L}{\delta x^{*}}, \frac{\delta _{\beta } L}{\delta u^{*}}\right) \) is constant on \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \).
Then \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is weakly sharp if and only if there exists \( \displaystyle \gamma > 0 \) such that
where \( \displaystyle d\left( (x,u), (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}\right) = \min _{(y,w) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}\parallel (x,u) - (y,w) \parallel \).
Proof
“\(\Longrightarrow \)” Consider \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is weakly sharp. Therefore, by Definition 4.1, there exists \( \displaystyle \gamma > 0 \) such that (1) (or (2)) is fulfilled. Further, taking into account the property of convexity for the solution set \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) associated to (CVIP) (see Remark 3.1), it follows \( \displaystyle proj_{(\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}(x,u) = (\hat{y}, \hat{w}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*},\; \forall (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) and, following Hiriart-Urruty and Lemaréchal [5], we obtain \( \displaystyle (x,u) - (\hat{y}, \hat{w}) \in T_{\mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}}(\hat{y}, \hat{w}) \cap N_{(\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}(\hat{y}, \hat{w}) \). By hypothesis and Lemma 4.1, we get
Since
by (3), we obtain \( \displaystyle R(x,u) \ge \gamma d((x,u), (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}), \; \forall (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \).
“\(\Longleftarrow \)” Consider there exists \( \displaystyle \gamma > 0 \) such that \( \displaystyle R(x,u) \ge \gamma d((x,u), (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}), \; \forall (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}. \) Obviously, for any \( \displaystyle (y,w) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \), the case \( \displaystyle T_{\mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}}(y,w) \cap N_{(\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}(y,w) = \left\{ (0,0) \right\} \) involves \( \displaystyle \left[ T_{\mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}}(y,w) \cap N_{(\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}(y,w)\right] ^{\circ } = \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \) and, consequently, \( \displaystyle \gamma B \subset \left( \frac{\delta _{\beta } L}{\delta y}, \frac{\delta _{\beta } L}{\delta w}\right) + \left[ T_{\mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}}(y,w) \cap N_{(\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}(y,w)\right] ^{\circ }, \; \forall (y,w) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is trivial. In the following, let \( \displaystyle (0,0) \ne (\overline{x}, \overline{u}) \in T_{\mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}}(y,w) \cap N_{(\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}(y,w) \) involving there exists a sequence \( \displaystyle (\overline{x}^{k}, \overline{u}^{k}) \) converging to \((\overline{x}, \overline{u})\) with \( \displaystyle (y,w) + t_{k}(\overline{x}^{k}, \overline{u}^{k})\in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \) (for some sequence of positive numbers \( \displaystyle \lbrace t_{k} \rbrace \) decreasing to zero), such that
where \( \displaystyle H_{\overline{x}, \overline{u}} = \left\{ (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}}:\langle (\overline{x}, \overline{u}); (x,u) - (y,w) \rangle = 0 \right\} \) is a hyperplane passing through (y, w) and orthogonal to \((\overline{x}, \overline{u})\). By hypothesis and (4), it results \( \displaystyle R((y,w) + t_{k}(\overline{x}^{k}, \overline{u}^{k})) \ge \gamma \frac{t_{k} \langle (\overline{x}, \overline{u}); (\overline{x}^{k}, \overline{u}^{k})\rangle }{\Vert (\overline{x}, \overline{u}) \Vert } \), or, equivalently (\( \displaystyle R(y,w) = 0, \; \forall (y,w) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}\)),
Further, by taking the limit for \( \displaystyle k \rightarrow \infty \) in (5) and using a classical result of functional analysis, we obtain
where \( \displaystyle \lambda > 0 \). By Definition 2.2, the inequality (6) becomes
Now, taking into account the hypothesis and (7), for any \( \displaystyle (b,\upsilon ) \in B \), it follows \( \displaystyle \langle \gamma (b,\upsilon ) - (\frac{\delta _{\beta } L}{\delta y}, \frac{\delta _{\beta } L}{\delta w}); (\overline{x}, \overline{u}) \rangle = \langle \gamma (b,\upsilon ); (\overline{x}, \overline{u}) \rangle - \langle (\frac{\delta _{\beta } R}{\delta y},\frac{\delta _{\beta } R}{\delta w}); (\overline{x}, \overline{u}) \rangle \le \gamma \Vert (\overline{x}, \overline{u}) \Vert - \gamma \Vert (\overline{x}, \overline{u}) \Vert = 0 \) and the proof is complete. \(\square \)
The second characterization result of weak sharpness for \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is based on the notion of minimum principle sufficiency property, introduced by Ferris and Mangasarian [4].
Definition 4.2
The controlled variational inequality (CVIP) satisfies minimum principle sufficiency property if \( \displaystyle \mathcal {A}(x^{*},u^{*}) = (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}, \; \forall (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \).
Lemma 4.2
The inclusion \( \displaystyle \arg \max _{(y,w) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}} \langle (x,u); (y,w) \rangle \subset (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is fulfilled for any \( \displaystyle (x,u) \in int\left( \bigcap _{(\overline{x}, \overline{u}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}\left[ T_{\mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}}(\overline{x}, \overline{u}) \cap N_{(\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}}(\overline{x}, \overline{u})\right] ^{\circ }\right) \ne \emptyset \).
Theorem 4.2
Let the solution set \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) associated to (CVIP) be weakly sharp and the path-independent curvilinear integral functional \( \displaystyle L(x,u) \) be convex on \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \). Then (CVIP) satisfies minimum principle sufficiency property.
Theorem 4.3
Consider the functional \( \displaystyle R(x,u) \) is differentiable on \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) and the path-independent curvilinear integral functional \( \displaystyle L(x,u) \) is convex on \( \displaystyle \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \). Also, for any \( \displaystyle (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}, \; (v,\mu ) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}}, \;(z,\nu ) \in \mathcal {Z}(x^{*},u^{*}) \), assume the following implication
is fulfilled and \( \displaystyle \left( \frac{\delta _{\beta } L}{\delta x^{*}}, \frac{\delta _{\beta } L}{\delta u^{*}}\right) \) is constant on \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \). Then (CVIP) satisfies minimum principle sufficiency property if and only if \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is weakly sharp.
Proof
“\(\Longrightarrow \)” Let (CVIP) satisfies minimum principle sufficiency property. In consequence, \( \displaystyle \mathcal {A}(x^{*},u^{*}) = (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \), for any \( \displaystyle (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \). Obviously, for \( \displaystyle (x^{*},u^{*}) \in (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) and \( \displaystyle (x, u) \in \overline{\mathcal {\mathbf {X}}} \times \overline{\mathcal {\mathbf {U}}} \), we obtain
Further, for \( \displaystyle P(x,u) = \langle (\frac{\delta _{\beta } L}{\delta x^{*}},\frac{\delta _{\beta } L}{\delta u^{*}}); (x,u) \rangle , \; (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \), we get \( \displaystyle \mathcal {A}(x^{*},u^{*}) \) the solution set for \( \displaystyle \min _{(x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}}P(x,u) \). For other related investigations, the readers are directed to Mangasarian and Meyer [7]. We can write \( \displaystyle P(x,u) - P(\tilde{x},\tilde{u}) \ge \gamma d((x,u), \mathcal {A}(x^{*},u^{*})), \; \forall (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}}, \; (\tilde{x}, \tilde{u}) \in \mathcal {A}(x^{*},u^{*}) \), or, \( \displaystyle \langle (\frac{\delta _{\beta } L}{\delta x^{*}}, \frac{\delta _{\beta } L}{\delta u^{*}}); (x,u) - (x^{*},u^{*}) \rangle \ge \gamma d((x,u), (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*}), \; \forall (x,u) \in \mathcal {\mathbf {X}} \times \mathcal {\mathbf {U}} \), or, equivalently,
By (8), (9) and Theorem 4.1, we get \( \displaystyle (\mathcal {\mathbf {X}}\times \mathcal {\mathbf {U}})^{*} \) is weakly sharp.
“\(\Longleftarrow \)” This is a consequence of Theorem 4.2. \(\square \)
References
Alshahrani, M., Al-Homidan S., Ansari, Q.H.: Minimum and maximum principle sufficiency properties for nonsmooth variational inequalities. Optim. Lett. 10, 805–819 (2016)
Burke, J.V., Ferris, M.C.: Weak sharp minima in mathematical programming. SIAM J. Control Optim. 31, 1340–1359 (1993)
Clarke, F.H.: Functional Analysis, Calculus of Variations and Optimal Control. Springer, London (2013)
Ferris, M.C., Mangasarian, O.L.: Minimum principle sufficiency. Math. Program. 57, 1–14 (1992)
Hiriart-Urruty, J.-B., Lemaréchal, C.: Fundamentals of Convex Analysis. Springer, Berlin (2001)
Liu, Y., Wu, Z.: Characterization of weakly sharp solutions of a variational inequality by its primal gap function. Optim. Lett. 10, 563–576 (2016)
Mangasarian, O.L., Meyer, R.R.: Nonlinear perturbation of linear programs. SIAM J. Control Optim. 17, 745–752 (1979)
Marcotte, P., Zhu, D.: Weak sharp solutions of variational inequalities. SIAM J. Optim. 9, 179–189 (1998)
Oveisiha, M., Zafarani, J.: Generalized Minty vector variational-like inequalities and vector optimization problems in Asplund spaces. Optim. Lett. 7, 709–721 (2013)
Patriksson, M.: A unified framework of descent algorithms for nonlinear programs and variational inequalities. Ph.D. thesis, Linköping Institute of Technology (1993)
Polyak, B.T.: Introduction to Optimization. Optimization Software. Publications Division, New York (1987)
Treanţă, S.: Multiobjective fractional variational problem on higher-order jet bundles. Commun. Math. Stat. 4, 323–340 (2016)
Treanţă, S.: Higher-order Hamilton dynamics and Hamilton-Jacobi divergence PDE. Comput. Math. Appl. 75, 547–560 (2018)
Treanţă, S., Arana-Jiménez, M.: On generalized KT-pseudoinvex control problems involving multiple integral functionals. Eur. J. Control 43, 39–45 (2018)
Wu, Z., Wu, S.Y.: Weak sharp solutions of variational inequalities in Hilbert spaces. SIAM J. Optim. 14, 1011–1027 (2004)
Zhu, S.K.: Weak sharp efficiency in multiobjective optimization. Optim. Lett. 10, 1287–1301 (2016)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Treanţă, S. (2020). On Controlled Variational Inequalities Involving Convex Functionals. In: Le Thi, H., Le, H., Pham Dinh, T. (eds) Optimization of Complex Systems: Theory, Models, Algorithms and Applications. WCGO 2019. Advances in Intelligent Systems and Computing, vol 991. Springer, Cham. https://doi.org/10.1007/978-3-030-21803-4_17
Download citation
DOI: https://doi.org/10.1007/978-3-030-21803-4_17
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-21802-7
Online ISBN: 978-3-030-21803-4
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)