Abstract
In this paper, we consider semi-infinite mathematical programming problems with equilibrium constraints (SIMPPEC). By using the notion of convexificators, we establish sufficient optimality conditions for the SIMPPEC. We formulate Wolfe and Mond–Weir-type dual models for the SIMPPEC under the invexity and generalized invexity assumptions. Weak and strong duality theorems are established to relate the SIMPPEC and two dual programs in the framework of convexificators.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
A semi-infinite programming (SIP) is an optimization problem in finitely many variables on a feasible set described by infinitely many constraints. There are many applications of SIP in various fields such as robust optimization, Chebyshev approximation, optimal control, robotics, transportation problems, mathematical physics, fuzzy sets, cooperative games, engineering design (see [1, 2]). For basic theory, survey articles on SIP we refer to [3] and for monograph [4].
The notion of convexificators can be seen as a generalization of subdifferentials. Jeyakumar et al. [5] have shown that the Clarke subdifferentials [6], Michel–Penot subdifferentials [7], Ioffe–Mordukhovich subdifferentials [8] and Treiman subdifferentials [9] of a locally Lipschitz real-valued function are convexificators and these known subdifferentials may contain the convex hull of a convexificator. Convexificators are not necessarily compact or convex, unlike some of the subdifferentials which are compact and convex objects. We refer to the recent results [10,11,12,13] and the references therein for more details related to the convexificators.
Usually, generalized convex functions have been introduced in order to weaken the convexity requirements as much as possible to obtain results related to optimization theory. One of the significant generalization of convex function is invex function [14, 15]. The class of invex functions preserves many properties of the class of convex functions and has shown to be very useful in a variety of applications [16]. It is well known that optimality and duality theory provides the foundation of algorithms for a solution of an optimization problem and hence constitutes an important portion in the study of mathematical programming. Wolfe [17] and Mond–Weir [18] dual models have been studied for semi-infinite programming problems [19, 20], mathematical programs with vanishing constraints [21], bi-level problems [22] and mathematical programming problem with equilibrium constraints [23].
The class of mathematical programs with equilibrium constraints is an extension of the class of bilevel programming problems [24,25,26,27], which is also known as mathematical programs with optimization constraints. Mathematical programming problem with equilibrium constraints (MPPEC) plays a vital role in many fields such as engineering design, capacity enhancement in traffic, economic equilibrium, dynamic pricing in telecommunication networks and multilevel games [28,29,30,31]. Many practical problems in these fields have been modeled using the MPPEC formulation. In practice, it is natural that an MPPEC may arise where infinitely many restrictions are present rather than finite many restrictions in finitely many variables. This gives us a motivation to formulate semi-infinite mathematical programming problem with equilibrium constraints (SIMPPEC).
In this paper, we establish sufficient optimality condition for the SIMPPEC and provide a supportive example corresponding to the sufficient optimality condition. We also derive the weak and strong duality theorems relating to the SIMPPEC and two dual models (Wolfe and Mond–Weir) under the framework of convexificators. The organization of this paper is as follows: In Sect. 2, we give some preliminaries, definitions, and results. In Sect. 3, we establish sufficient optimality condition for the SIMPPEC using convexificators. In Sect. 4, we derive weak and strong duality theorems relating to the SIMPPEC and two dual models using \(\partial ^*\)-invex, \(\partial ^*\)-pseudoinvex and \(\partial ^*\)-quasiinvex functions. In Sect. 5, we conclude the results of this paper.
2 Preliminaries
Throughout the paper, \(\mathbb {R}^n\) denotes the n-dimensional Euclidean space. Let C be a nonempty subset of \(\mathbb {R}^n\). The convex hull of C and the convex cone generated by C are denoted by \(co \ C\) and \(cone \ C\), respectively.
The negative polar cone is defined by \( C^- = \{u \in \mathbb {R}^n \vert \ \forall \ x \in C, \ \langle x,u \rangle \leqslant 0\}.\)
Let C be a nonempty subset of \(\mathbb {R}^n\) and \(x \in cl \ C \) (closure of C), then the contingent cone T(x, C) to C at x is defined by
We consider SIMPPEC in the following form:
where the index set T is an infinite compact subset of \(\mathbb {R}^n, \ F: \mathbb {R}^n \rightarrow \mathbb {R}, \ g : \mathbb {R}^n \times T \rightarrow \mathbb {R},\ h: \mathbb {R}^n \rightarrow \mathbb {R}^p,\ \Phi : \mathbb {R}^n \rightarrow \mathbb {R}^l \ \text {and} \ \Psi : \mathbb {R}^n \rightarrow \mathbb {R}^l\) are given functions.
Along the lines of [19] for a given feasible vector \(\tilde{x} \) for the SIMPPEC, we define the following index sets:
where the set \(\omega \) is known as degenerate set.
Definition 2.1
Let \(F:\mathbb {R}^n\rightarrow \mathbb {R}\cup \left\{ +\infty \right\} \) be an extended real-valued function, \(x\in \mathbb {R}^n\) and let F(x) be finite. Then, the lower and upper Dini directional derivatives of F at x in the direction v are defined, respectively, by
and
Definition 2.2
(see [5]) A function \(F: \mathbb {R}^n \rightarrow \mathbb {R} \cup \{+ \infty \}\) is said to admit an upper convexificator, \(\partial ^* F(x)\) at \(x \in \mathbb {R}^n\) if \(\partial ^* F(x) \subseteq \mathbb {R}^n\) is a closed set and, for every \(v \in \mathbb {R}^n,\)
Definition 2.3
(see [5]) A function \(F: \mathbb {R}^n \rightarrow \mathbb {R} \cup \{+ \infty \}\) is said to admit a lower convexificator, \(\partial _* F(x)\) at \(x \in \mathbb {R}^n\) if \(\partial _* F(x) \subseteq \mathbb {R}^n\) is a closed set and, for every \(v \in \mathbb {R}^n,\)
The function F is said to have a convexificator \(\partial ^*F(x)\subseteq \mathbb {R}^n\) at \(x\in \mathbb {R}^n,\) if \(\partial ^*F(x)\) is both upper and lower convexificator of F at x.
Definition 2.4
(see [32]) A function \(F: \mathbb {R}^n \rightarrow \mathbb {R} \cup \{+\infty \}\) is said to admit an upper semi-regular convexificator, \(\partial ^* F(x)\) at \( x \in \mathbb {R}^n\) if \(\partial ^* F(x) \subseteq \mathbb {R}^n\) is a closed set and, for every \(v \in \mathbb {R}^n\)
If equality holds in (2.1), then \(\partial ^* F(x)\) is called an upper regular convexificator of F at x.
Based on the definitions of invex function [16] and generalized invex functions [33], we will introduce the definitions of invex function and generalized invex functions in terms of convexificators.
Definition 2.5
Let \(\eta :\mathbb {R}^n \times \mathbb {R}^n \rightarrow \mathbb {R}^n\) be a kernel function and let \(F: \mathbb {R}^n \rightarrow \mathbb {R} \cup \{+ \infty \}\) be an extended real-valued function, which admit convexificator at \(\tilde{z} \in \mathbb {R}^n\). Then F is said to be
-
(i)
\(\partial ^*\)-invex at \(\tilde{z}\) with respect to \(\eta \) if for every \(x \in \mathbb {R}^n,\)
$$\begin{aligned} F(x) \geqslant F(\tilde{z}) + \langle \xi , \eta (x, \tilde{z}) \rangle , \forall \ \xi \in \partial ^*F(\tilde{z}). \end{aligned}$$ -
(ii)
\(\partial ^*\)-pseudoinvex at \( \tilde{z}\) with respect to \(\eta \) if for every \(x \in \mathbb {R}^n\),
$$\begin{aligned} \exists \ \xi \in \partial ^*F(\tilde{z}), \ \langle \xi , \eta (x, \tilde{z}) \rangle \geqslant 0 \Rightarrow F(x) \geqslant F(\tilde{z}). \end{aligned}$$ -
(iii)
\(\partial ^*\)-quasiinvex at \( \tilde{z}\) with respect to \(\eta \) if for every \(x \in \mathbb {R}^n\),
$$\begin{aligned} F(x) \leqslant F(\tilde{z}) \Rightarrow \langle \xi , \eta (x, \tilde{z}) \rangle \leqslant 0, \forall \ \xi \in \partial ^*F(\tilde{z}). \end{aligned}$$
Remark 1
Based on Definition 2.5, the definition of \(\partial ^*\)-invex function and generalized \(\partial ^*\)-invex functions can also be given in terms of upper semi-regular convexificators.
Pandey and Mishra [34] presented the following notations for SIMPPEC given for MPPEC by Ardali et al. [35]:
where \(t_1, t_2, \cdots , t_{m} \in T_g(\tilde{z}), \ m \leqslant n+1,\) and \(\tilde{z}\) is a feasible point of SIMPPEC.
The following definitions are taken from Pandey and Mishra [34] for SIMPPEC.
Definition 2.6
Let \(\tilde{z}\) be a feasible point of SIMPPEC, and assume that all functions have convexificators considered above at \(\tilde{z}.\) We say that the generalized standard Abadie constraint qualification (GS Abadie CQ) holds at \(\tilde{z}\) if at least one of the dual sets used in the definition of \(\Gamma (\tilde{z})\) is nonzero and
Definition 2.7
A feasible point \(\tilde{z}\) of SIMPPEC is called the generalized alternatively stationary (GA-stationary) point if there exist \(\tau = (\tau ^g, \tau ^h, \tau ^{\Phi }, \tau ^{\Psi }) \in \mathbb {R}^{k+p+2l}, \gamma \in (\gamma ^h, \gamma ^{\Phi }, \gamma ^{\Psi }) \in \mathbb {R}^{p+2l}\) and \(t_1, t_2, \cdots , t_m \in T_g(\tilde{z}),\ m \leqslant n+1\), such that the following conditions hold:
Definition 2.8
A feasible point \(\tilde{z}\) of SIMPPEC is called the generalized strong stationary (GS-stationary) point if there exist \(\tau = (\tau ^g, \tau ^h, \tau ^{\Phi }, \tau ^{\Psi }) \in \mathbb {R}^{k+p+2l}, \gamma \in (\gamma ^h, \gamma ^{\Phi }, \gamma ^{\Psi }) \in \mathbb {R}^{p+2l}\) and \(t_1, t_2, \cdots , t_m \in T_g(\tilde{z}),\ m \leqslant n+1\), satisfying conditions (2.2) and (2.3) together with the following conditions:
Note The following index sets will be used in Sects. 3 and 4, respectively:
In the next section, we will obtain sufficient optimality condition under generalized invexity assumptions using the notion of convexificators.
3 Optimality Condition
The following result shows that GS-stationarity is a necessary optimality condition for SIMPPEC.
Theorem 3.1
[34] Let \( \tilde{z}\) be a local optimal solution of SIMPPEC. Suppose that F is locally Lipschitz function at \(\tilde{z}\), which admits a bounded upper semi-regular convexificator \(\partial ^*F(\tilde{z})\). Assume also that GS-ACQ holds at \(\tilde{z}\) and that the cone
is closed, then \(\tilde{z}\) is a GS-stationary point.
Corollary 3.1
[34] Let \(\tilde{z}\) be a local optimal solution of SIMPPEC. Suppose that F is locally Lipschitz near \(\tilde{z}\). Assume also that F and effective constraint functions admit bounded upper semi-regular convexificators at \(\tilde{z}\). If GS-ACQ holds at \(\tilde{z}\), then \(\tilde{z}\) is a GS-stationary point.
The following theorem shows that under generalized \(\partial ^*\)-invexity assumptions, GA-stationarity turns into a global sufficient optimality condition.
Theorem 3.2
Let \(\tilde{z}\) be a feasible GA-stationary point of SIMPPEC and assume that F is \(\partial ^*\)-pseudoinvex at \(\tilde{z}\) with respect to the kernel \(\eta \) and \(g(.,t) \ (t \in T_g), \pm h_r\ (r= 1,2, \cdots , p), \ -\Phi _i\ (i \in \delta \cup \omega ), -\Psi _i\ (i \in \omega \cup \kappa ) \) are \(\partial ^*\)-quasiinvex at \(\tilde{z}\) with respect to the common kernel \(\eta .\) If \(\omega _{\gamma }^{\Phi } \cup \omega _{\gamma }^{\Psi } \cup \delta _{\gamma }^+ \cup \kappa _{\gamma }^+ = \varnothing \ then \ \tilde{z}\) is a global optimal solution of SIMPPEC.
Proof
Let us consider that x be any arbitrary feasible point of SIMPPEC. Then for any \(t_i \in T_g(\tilde{z})\).
by \(\partial ^*\)-quasiinvexity of \(g(x,t_i) \) at \(\tilde{z}\), it follows that
Similarly, we have
If \(\omega _\gamma ^{\Phi } \cup \omega _\gamma ^{\Psi } \cup \delta _\gamma ^+ \cup \kappa _\gamma ^+ = \varnothing ,\) multiplying (3.1)–(3.5) by \(\tau _i^g \geqslant 0 \, (i = 1,2, \cdots , m), \ \tau _r^h> 0 \, (r= 1,2,\cdots , p), \, \gamma _r^h> 0 \, (r=1,2,\cdots ,p), \, \tau _i^{\Phi }> 0 \ (i \in \delta \cup \omega ), \, \tau _i^{\Psi } > 0 \ (i \in \omega \cup \kappa ),\) respectively, and adding, we obtain
Therefore,
for all \( {\xi }_i^g \in co \partial ^*g(\tilde{z},t_i), \ {\zeta _r} \in co \partial ^* h_r(\tilde{z}) , \ {\nu }_r \in co \partial ^*(-h_r)(\tilde{z}), \ {\xi }_i^{\Phi } \in co \partial ^*(-\Phi _i)(\tilde{z}) \) and \( {\xi }_i^{\Psi } \in co \partial ^*(-\Psi _i)(\tilde{z}). \) Thus by GA-stationarity of \(\tilde{z}\), we can choose \(\xi \in co \partial ^*F(\tilde{z}),\) such that
By \(\partial ^*\)-pseudoinvexity of F at \(\tilde{z}\) with respect to the common kernel \(\eta \), we have \(F(x) \geqslant F(\tilde{z})\) for all feasible points x. Hence \(\tilde{z}\) is a global optimal solution of SIMPPEC. This completes the proof.
The following example illustrates Theorem 3.2.
Example 3.1
Consider the following SIMPPEC:
SIMPPEC
Here F is \(\partial ^*\)-pseudoinvex at \(\tilde{z} = 0\) with respect to the kernel, \( \eta (x, \tilde{z})=e^x \sin \tilde{z}.\) Further, \(g, -\theta \ \text {and} -\Psi \) are \(\partial ^*\)-quasiinvex at \(\tilde{z} = 0\) with respect to the common kernel, \(\eta (x, \tilde{z}) = e^x \sin \tilde{z}\). The feasible point for the given SIMPPEC is \(\tilde{z} = 0.\) We have \( co \partial ^*F(0) = [-\pi ,\pi ], \ co \partial ^*g(0, t_1) = \{0\}, \ t_1 = 0,\)\(co \partial ^*(-\theta )(0) = [-1,1] \) and \( co \partial ^*(-\Psi )(0) = \{0\}.\) One can easily verify that there exist \(\tau ^g= 1, \tau ^{\theta }=1\) and \(\tau ^{\Psi } = 1\) such that \(\tilde{z} = 0\) is a GA-stationary point and \(\tilde{z} = 0\) is a global optimal solution for the given primal SIMPPEC. Hence the assumptions of the Theorem 3.2, are satisfied.
4 Duality
In this section, we formulate and study a Wolfe-type dual problem for SIMPPEC using \(\partial ^*\)-invexity. We also formulate Mond–Weir-type dual problem and study SIMPPEC using \(\partial ^*\)-invexity and generalized \(\partial ^*\)-invexity assumptions.
The Wolfe-type dual for SIMPPEC is formulated as follows:
s.t.
where \(\rho _r^h= \tau _r^h - \gamma _r^h, \tau = (\tau ^g, \tau ^h, \tau ^{\Phi }, \tau ^{\Psi }) \in \mathbb {R}^{k+p+2l}\), \( \gamma = (\gamma ^h, \gamma ^{\Phi }, \gamma ^{\Psi }) \in \mathbb {R}^{p+2l}\) and \(t_1, t_2, \cdots , t_m \in T_g,\ m \leqslant n + 1.\)
Theorem 4.1
(Weak Duality) Let \(\tilde{x}\) be feasible for SIMPPEC, \((z, \tau )\) be feasible for the dual WD and the index sets \(T_g, \delta , \omega , \kappa \) are defined accordingly. Suppose that \(F, \ g(., t) \ (t \in T), \ \pm h_r \ (r=1,2,\cdots ,p), \ -\Phi _i \ (i \in \delta \cup \omega ), \ -\Psi _i \ (i \in \omega \cup \kappa )\) admit bounded upper semi-regular convexificators and are \(\partial ^*\)-invex functions at z, with respect to the common kernel \(\eta .\) If \(\omega _\gamma ^{\Phi } \cup \omega _\gamma ^{\Psi } \cup \delta _\gamma ^+ \cup \kappa _\gamma ^+ = \varnothing ,\) then for any x feasible for the SIMPPEC, we have
Proof
Suppose that x be any arbitrary feasible point for the SIMPPEC. It follows that
Since F is invex at z, with respect to the kernel \(\eta \), we have
Similarly,
If \(\omega _\gamma ^{\Phi } \cup \omega _\gamma ^{\Psi } \cup \delta _\gamma ^+ \cup \kappa _\gamma ^+ = \varnothing ,\) then multiplying (4.3)–(4.7) by \(\tau _i^g \geqslant 0 \ (i = 1,2, \cdots , m), \ \tau _r^h> 0 \ (r= 1,2,\cdots , p), \ \gamma _r^h> 0 \ (r=1,2,\cdots ,p), \ \tau _i^{\Phi }> 0 \ (i \in \delta \cup \omega ), \ \tau _i^{\Psi } > 0 \ (i \in \omega \cup \kappa ),\) respectively, and adding (4.2)–(4.7), it follows that
From (4.1), \( \exists \ \tilde{\xi } \in co \partial ^*F(z), \ \tilde{\xi }_i^g \in co \partial ^*g(z,t_i)\ (t_i \in T_g), \ \tilde{\zeta _r} \in co \partial ^* h_r(z) , \ \tilde{\nu }_r \in co \partial ^*(-h_r)(z), \ \tilde{\xi }_i^{\Phi } \in co \partial ^*(-\Phi _i)(z) \ \text {and} \ \tilde{\xi }_i^{\Psi } \in co \partial ^*(-\Psi _i)(z), \) such that
Therefore,
Using the feasibility of x for SIMPPEC, i.e., \(g(x,t_i) \leqslant 0, \ h_r(x) = 0, \ \Phi _i(x) \geqslant 0, \ \Psi _i(x) \geqslant 0\), it follows that
Hence,
where \(\rho _r^h = \tau _r^h - \gamma _r^h.\) This completes the proof.
Theorem 4.2
(Strong Duality) Let \(\tilde{x}\) be a local optimal solution of SIMPPEC and let F be locally Lipschitz near \(\tilde{x}\). Suppose that \(F, \ g(.,t) \ (t \in T), \ \pm h_r(r= 1,2,\cdots , p), -\Phi _i(i \in \delta \cup \omega ), -\Psi _i(i \in \omega \cup \kappa )\) admit bounded upper semi-regular convexificators and are \(\partial ^*\)-invex functions at \(\tilde{x}\) with respect to the common kernel \(\eta \). If GS-ACQ holds at \( \tilde{x},\) then there exists \(\tilde{\tau } = (\tilde{\tau }^g, \tilde{\tau }^h, \tilde{\tau }^{\Phi }, \tilde{\tau }^{\Psi }) \in \mathbb {R}^{k+p+2l},\) such that \((\tilde{x}, \tilde{\tau })\) is an optimal solution of the dual WD and the respective objective values are equal.
Proof
Since, \(\tilde{x}\) is a local optimal solution of SIMPPEC and GS-ACQ is satisfied at \(\tilde{x},\) now, using Corollary 3.1, \( \exists \ \tilde{\tau } = (\tilde{\tau }^g, \tilde{\tau }^h, \tilde{\tau }^{\Phi }, \tilde{\tau }^{\Psi }) \in \mathbb {R}^{k+p+2l}, \tilde{\gamma } \in (\tilde{\gamma }^h,\tilde{\gamma }^{\Phi },\tilde{\gamma }^{\Psi }) \in \mathbb {R}^{p+2l},\) and indices \(t_1, t_2, \cdots , t_m \in T_g(\tilde{x}),\ m \leqslant n + 1,\) such that GS-stationarity conditions for SIMPPEC are satisfied, that is, \( \exists \ \tilde{\xi } \in co \partial ^* F(\tilde{x}), \ \tilde{\xi }_i^g \in co \partial ^* g(\tilde{x},t_i), \ \tilde{\zeta }_r \in co \partial ^* h_r(\tilde{x}), \ \tilde{\nu }_r \in co \partial ^*(-h_r)(\tilde{x}), \ \tilde{\xi }_i^{\Phi } \in co \partial ^*(-\Phi _i)(\tilde{x}) \ \text {and} \ \tilde{\xi }_i^{\Psi } \in co \partial ^*(-\Psi _i)(\tilde{x}),\) such that
Therefore \((\tilde{x}, \tilde{\tau })\) is feasible for the dual WD. Now, using Theorem 4.1, we obtain
where \(\rho _r^h = \tau _r^h - \gamma _r^h,\) for any feasible solution \((z, \tau )\) for the dual WD. Using the feasibility condition of SIMPPEC and dual WD, that is, for \(t_i \in T_g(\tilde{x}), \ g(\tilde{x}, t_i) = 0,\ h_r(\tilde{x}) = 0, (r= 1,2, \cdots ,p), \ \Phi _i(\tilde{x})=0, \forall i \in \delta \cup \omega , \ \text {and} \ \Psi _i(\tilde{x}) = 0, \forall i \in \omega \cup \kappa ,\) then, we have
where \(\tilde{\rho }_r^h = \tilde{\tau }_r^h - \tilde{\gamma }_r^h.\) Using (4.8) and (4.9), it follows that
Hence, \((\tilde{x}, \tilde{\tau })\) is an optimal solution for the dual WD and the respective objective values are equal. This completes the proof.
Now, we formulate the Mond–Weir-type dual problem (MWDP) for SIMPPEC and establish duality theorems using convexificators.
s.t.
where \(\tau = (\tau ^g, \tau ^h, \tau ^{\Phi }, \tau ^{\Psi }) \in \mathbb {R}^{k+p+2l}, \ \gamma = (\gamma ^h, \gamma ^{\Phi }, \gamma ^{\Psi }) \in \mathbb {R}^{p+2l} \ \text {and} \ \ t_1, t_2, \cdots , t_m \in T_g(\tilde{x}),\ m \leqslant n + 1.\)
Theorem 4.3
(Weak Duality) Let \(\tilde{x}\) be feasible for SIMPPEC, \((z, \tau )\) be feasible for the MWDP and the index sets \(T_g, \delta , \omega , \kappa \) be defined accordingly. Suppose that \(F, g(.,t) \ (t \in T), \pm h_r(r=1,2,\cdots ,p), -\Phi _i(i \in \delta \cup \omega ), -\Psi _i (i \in \omega \cup \kappa )\) admit bounded upper semi-regular convexificators and are \(\partial ^*\)-invex functions at z, with respect to the common kernel \(\eta .\) If \(\omega _\gamma ^{\Phi } \cup \omega _\gamma ^{\Psi } \cup \delta _\gamma ^+ \cup \kappa _\gamma ^+ = \varnothing ,\) then for any x feasible for SIMPPEC, we have
Proof
Since F is invex at z, with respect to the kernel \(\eta ,\) then, we have
Similarly, we have
If \(\omega _\gamma ^{\Phi } \cup \omega _\gamma ^{\Psi } \cup \delta _\gamma ^+ \cup \kappa _\gamma ^+ = \varnothing ,\) multiplying (4.12)–(4.16) by \(\tau _i^g \geqslant 0 \ (i = 1,2, \cdots , m), \tau _r^h> 0 \ (r= 1,2,\cdots , p), \ \gamma _r^h> 0 \ (r=1,2,\cdots ,p), \ \tau _i^{\Phi }> 0 \ (i \in \delta \cup \omega ), \ \tau _i^{\Psi } > 0 \ (i \in \omega \cup \kappa ),\) respectively, and adding (4.11)–(4.16), we obtain
From (4.10), \( \exists \ \tilde{\xi } \in co \partial ^*F(z), \ \tilde{\xi }_i^g \in co \partial ^*g(z,t_i), \tilde{\zeta }_r \in co \partial ^* h_r(z), \ \tilde{\nu }_r \in co \partial ^*(-h_r)(z), \ \tilde{\xi }_i^{\Phi } \in co \partial ^*(-\Phi _i)(z) \ \text {and} \ \tilde{\xi }_i^{\Psi } \in co \partial ^*(-\Psi _i)(z), \) such that
Therefore,
Using the feasibility of x and z for SIMPPEC and MWDP, respectively, we obtain
This completes the proof.
Theorem 4.4
(Strong Duality) Let \(\tilde{x}\) be a local optimal solution of SIMPPEC and let F be locally Lipschitz near \(\tilde{x}\). Suppose that \(F, \ g(.,t) \ (t \in T), \ \pm h_r \ (r= 1,2,\cdots , p), \ -\Phi _i \ (i \in \delta \cup \omega ), \ -\Psi _i \ (i \in \omega \cup \kappa )\) admit bounded upper semi-regular convexificators and are \(\partial ^*\)-invex functions at \(\tilde{x}\) with respect to the common kernel \(\eta .\) If GS-ACQ holds at \( \tilde{x},\) then there exists \(\tilde{\tau }\), such that \((\tilde{x}, \tilde{\tau })\) is an optimal solution of the MWDP and the respective objective values are equal.
Proof
Since, \(\tilde{x}\) is a local optimal solution of SIMPPEC and the GS-ACQ is satisfied at \(\tilde{x},\) now using Corollary 3.1, \( \exists \ \tilde{\tau } = (\tilde{\tau }^g, \tilde{\tau }^h, \tilde{\tau }^{\Phi }, \tilde{\tau }^{\Psi }) \in \mathbb {R}^{k+p+2l}, \tilde{\gamma } \in (\tilde{\gamma }^h,\tilde{\gamma }^{\Phi },\tilde{\gamma }^{\Psi }) \in \mathbb {R}^{p+2l},\) and indices \(t_1, t_2, \cdots , t_m \in T_g(\tilde{x}), \ m \leqslant n + 1,\) such that the GS-stationarity conditions for SIMPPEC are satisfied, that is, \( \exists \ \tilde{\xi } \in co \partial ^* F(\tilde{x}), \ \tilde{\xi }_i^g \in co \partial ^* g(\tilde{x},t_i), \ \tilde{\zeta }_r \in co \partial ^* h_r(\tilde{x}), \ \tilde{\nu }_r \in co \partial ^*(-h_r)(\tilde{x}), \ \tilde{\xi }_i^{\Phi } \in co \partial ^*(-\Phi _i)(\tilde{x}) \ \text {and} \ \tilde{\xi }_i^{\Psi } \in co \partial ^*(-\Psi _i)(\tilde{x}),\) such that
Since \(\tilde{x}\) is an optimal solution for SIMPPEC, we have
Therefore \((\tilde{x}, \tilde{\tau })\) is feasible for MWDP. By Theorem 4.3, for any feasible \((z, \tau ),\) we have
It follows that \((\tilde{x}, \tilde{\tau })\) is an optimal solution for MWDP and the respective objective values are equal. This completes the proof.
Now, we establish weak and strong duality theorems for SIMPPEC and its MWDP under generalized \(\partial ^*\)-invexity assumptions.
Theorem 4.5
(Weak Duality) Let \(\tilde{x}\) be feasible for SIMPPEC, \((z, \tau )\) be feasible for the MWDP and the index sets \(T_g, \delta , \omega , \kappa \) are defined accordingly. Suppose that F is \(\partial ^*\)-pseudoinvex at z, with respect to the kernel \(\eta \) and \(g(.,t) \ (t \in T), \ \pm h_r \ (r=1,2,\cdots ,p), \ -\Phi _i \ (i \in \delta \cup \omega ), \ -\Psi _i \ (i \in \omega \cup \kappa )\) admit bounded upper semi-regular convexificators and are \(\partial ^*\)-quasiinvex functions at z, with respect to the common kernel \(\eta .\) If \(\omega _\gamma ^{\Phi } \cup \omega _\gamma ^{\Psi } \cup \delta _\gamma ^+ \cup \kappa _\gamma ^+ = \varnothing ,\) then for any x feasible for the problem SIMPPEC, we have
Proof
Suppose that for some feasible point x, such that \(F(x) < F(z),\) then by \(\partial ^*\)-pseudoinvexity of F at z, with respect to the kernel \(\eta ,\) we have
From (4.10), \( \exists \, \tilde{\xi } \in co \partial ^*F(z), \, \tilde{\xi }_i^g \in co \partial ^*g(z,t_i), \tilde{\zeta }_r \in co \partial ^* h_r(z), \, \tilde{\nu }_r \in co \partial ^*(-h_r)(z),\) \(\tilde{\xi }_i^{\Phi } \in co \partial ^*(-\Phi _i)(z) \, \text {and} \, \tilde{\xi }_i^{\Psi } \in co \partial ^*(-\Psi _i)(z), \) such that
By (4.17), we have
For each \(t_i \in T_g, \ g(x,t_i) \leqslant 0 \leqslant g(z,t_i)\). Hence, by \(\partial ^*\)-quasiinvexity, it follows that
Similarly, we have
For any feasible point z of the MWDP, and for every \(r, -h_r(z) = h_r(x) = 0\). On the other hand, \(-\Phi _i(x) \leqslant -\Phi _i(z), \forall i \in \delta \cup \omega \), and \(-\Psi _i(x) \leqslant -\Psi _i(z), \forall i \in \omega \cup \kappa \). By \(\partial ^*\)-quasiinvexity, we have
From Eqs. (4.20)–(4.24), we have
Since \(\omega _\gamma ^{\Phi } \cup \omega _\gamma ^{\Psi } \cup \delta _\gamma ^+ \cup \kappa _\gamma ^+ = \varnothing ,\) we have
Therefore,
which contradicts (4.19). Hence \(F(x) \geqslant F(z)\). This completes the proof.
Theorem 4.6
(Strong Duality) Let \(\tilde{x}\) be a local optimal solution of SIMPPEC and let F be locally Lipschitz near \(\tilde{x}\). Suppose that F is \(\partial ^*\)-pseudoinvex at \(\tilde{x},\) with respect to the kernel \(\eta ,\) further \(g(.,t) \ (t \in T), \ \pm h_r \ (r= 1,2,\cdots , p), \ -\Phi _i \ (i \in \delta \cup \omega ),\ -\Psi _i\ (i \in \omega \cup \kappa )\) admit bounded upper semi-regular convexificators and are \(\partial ^*\)-quasiinvex functions at \(\tilde{x}\) with respect to the common kernel \(\eta .\) If GS-ACQ holds at \( \tilde{x},\) then there exists \(\tilde{\tau }, \) such that \((\tilde{x}, \tilde{\tau })\) is an optimal solution of the MWDP and the respective objective values are equal.
Proof
The proof follows on the lines of the proof of Theorem 4.4 by using Theorem 4.5.
5 Conclusions
We have studied SIMPPEC and established sufficient optimality condition under generalized invexity assumptions. We have introduced Wolfe and Mond–Weir-type dual models for the SIMPPEC in the framework of convexificators. We have established weak and strong duality theorems relating to the SIMPPEC and two dual models using \(\partial ^*\)-invexity, \(\partial ^*\)-pseudoinvexity and \(\partial ^*\)-quasiinvexity assumptions.
References
Polak, E.: On the mathematical foundation of nondifferentiable optimization in engineering design. SIAM Rev. 29, 21–89 (1987)
Hettich, R., Kortanek, K.O.: Semi-infinite programming: theory, methods and applications. SIAM Rev. 35, 380–429 (1993)
López, M.A., Still, G.: Semi-infinite programming. Eur. J. Oper. Res. 180(1), 491–518 (2007)
Goberna, M.A., López, M.A.: Linear Semi-infinite Optimization. Wiley, Chichester (1998)
Jeyakumar, V., Luc, D.T.: Nonsmooth calculus, minimality, and monotonicity of convexificators. J. Optim. Theory Appl. 101(3), 599–621 (1999)
Clarke, F.H.: Optimization and Nonsmooth Analysis. Wiley, New York, NY (1983)
Michel, P., Penot, J.P.: A generalized derivative for calm and stable functions. Differ. Integral Equ. 5, 433–454 (1992)
Ioffe, A.D.: Approximate subdifferentials and applications, II. Mathematika 33, 111–128 (1986)
Treiman, J.S.: The linear nonconvex generalized gradient and Lagrange multipliers. SIAM J. Optim. 5, 670–680 (1995)
Kabgani, A., Soleimani-damaneh, M., Zamani, M.: Optimality conditions in optimization problems with convex feasible set using convexificators. Math. Methods Oper. Res. 86(1), 103–121 (2017)
Laha, V., Mishra, S.K.: On vector optimization problems and vector variational inequalities using convexificators. Optimization 66(11), 1837–1850 (2017)
Kabgani, A., Soleimani-damaneh, M.: Characterization of (weakly/properly/robust) efficient solutions in nonsmooth semi-infinite multiobjective optimization using convexificators. Optimization 67(2), 217–235 (2018)
Alavi Hejazi, M., Movahedian, N., Nobakhtian, S.: Multiobjective problems: enhanced necessary conditions and new constraint qualifications through convexificators. Numer. Funct. Anal. Optim. 39(1), 11–37 (2018)
Hanson, M.A.: On sufficiency of the Kuhn–Tucker conditions. J. Math. Anal. Appl. 80, 545–550 (1981)
Craven, B.D.: Invex function and constrained local minima. Bull. Aust. Math. Soc. 24, 357–366 (1981)
Mishra, S.K., Giorgi, G.: Invexity and Optimization. Nonconvex Optimization and Its Applications, vol. 88. Springer, Berlin (2008)
Wolfe, P.: A duality theorem for nonlinear programming. Q. J. Appl. Math. 19, 239–244 (1961)
Mond, B., Weir, T.: Generalized Concavity and Duality, Generalized Concavity in Optimization and Economics. Academic Press, New York (1981)
Mishra, S.K., Jaiswal, M., An, L.T.H.: Duality for nonsmooth semi-infinite programming problems. Optim. Lett. 6, 261–271 (2012)
Pandey, Yogendra, Mishra, S.K.: On strong KKT type sufficient optimality conditions for nonsmooth multiobjective semi-infinite mathematical programming problems with equilibrium constraints. Oper. Res. Lett. 44(1), 148–151 (2016)
Mishra, S.K., Singh, V., Laha, V.: On duality for mathematical programs with vanishing constraints. Ann. Oper. Res. 243, 249–272 (2016)
Suneja, S.K., Kohli, B.: Optimality and duality results for bilevel programming problem using convexifactors. J. Optim. Theory Appl. 150, 1–19 (2011)
Pandey, Y., Mishra, S.K.: Duality for nonsmooth mathematical programming problems with equilibrium constraints using convexificators. J. Optim. Theory Appl. 171, 694–707 (2016)
Suh, S., Kim, T.J.: Solving nonlinear bilevel programming models of the equilibrium network design problem: a comparative review. Ann. Oper. Res. 34, 203–218 (1992)
Luo, Z.Q., Pang, J.S., Ralph, D.: Mathematical Programs with Equilibrium Constraints. Cambridge University Press, Cambridge (1996)
Colson, B., Marcotte, P., Savard, G.: A overview of bilevel optimization. Ann. Oper. Res. 153, 235–256 (2007)
Dempe, S., Zemkoho, A.B.: Bilevel road pricing: theoretical analysis and optimality conditions. Ann. Oper. Res. 196, 223–240 (2012)
Raghunathan, A.U., Biegler, L.T.: Mathematical programs with equilibrium constraints in process engineering. Comput. Chem. Eng. 27, 1381–1392 (2003)
Britz, W., Ferris, M., Kuhn, A.: Modeling water allocating institutions based on multiple optimization problems with equilibrium constraints. Environ. Model. Softw. 46, 196–207 (2013)
Harker, P.T., Pang, J.S.: Finite-dimensional variational inequality and nonlinear complementarity problems: a survey of theory, algorithms and applications. Math. Program. 48(1–3), 161–220 (1990)
Haslinger, J., Neittaanmäki, P.: Finite Element Approximation for Optimal Shape Design: Theory and Applications, p. xii. Wiley, Chichester (1988)
Dutta, J., Chandra, S.: Convexificators, generalized convexity and vector optimization. Optimization 53, 77–94 (2004)
Soleimani-damaneh, M.: Characterizations and applications of generalized invexity and monotonicity in Asplund spaces. Top 20(3), 592–613 (2012)
Pandey, Y., Mishra, S.K.: Optimality conditions and duality for semi-infinite mathematical programming problems with equilibrium constraints, using convexificators. Ann. Oper. Res. 269, 1–16 (2017)
Ansari Ardali, A., Movahedian, N., Nobakhtian, S.: Optimality conditions for nonsmooth mathematical programs with equilibrium constraints, using convexificators. Optimization 65(1), 67–85 (2016)
Acknowledgements
The authors are thankful to the anonymous referees for their valuable comments and suggestions which helped to improve the presentation of the paper.
Author information
Authors and Affiliations
Corresponding author
Additional information
The research of Shashi Kant Mishra was supported by Department of Science and Technology-Science and Engineering Research Board (No. MTR/2018/000121), India.
Rights and permissions
About this article
Cite this article
Joshi, B.C., Mishra, S.K. & Kumar, P. On Semi-infinite Mathematical Programming Problems with Equilibrium Constraints Using Generalized Convexity. J. Oper. Res. Soc. China 8, 619–636 (2020). https://doi.org/10.1007/s40305-019-00263-y
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40305-019-00263-y