Abstract
By applying the Brouwer fixed point theorem, we prove an existence result of solutions for strongly pseudomonotone quasi-variational inequalities which extends an analogous result in Kocvara and Outrata (Optim Methods Softw 5:275–295, 1995). The result is based on a new result concerning continuity property of solutions to a parametric variational inequality. Examples are given to illustrate our results.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction and preliminaries
Let F be a single-valued mapping from \(\mathbb {R}^n\) to \(\mathbb {R}^n\) and let K be a set-valued mapping from \(\mathbb {R}^n\) to \(\mathbb {R}^n\) such that K(x) is a closed convex set in \(\mathbb {R}^n\) for each \(x\in \mathbb {R}^n\). We consider the quasi-variational inequality (in short, QVI): Find \(x^*\in K(x^*)\) such that
Here, \(\langle \cdot ,\cdot \rangle \) denotes the inner product in \(\mathbb {R}^n\). When \(K(x)= K\), a nonempty closed convex subset of \(\mathbb {R}^n\), for all \(x\in \mathbb {R}^n\), the quasi-variational inequality (1.1) reduces to the classical variational inequality (in short, VIP): Find \(x^*\in K\) such that
Quasi-variational inequalities were introduced by Bensoussan and Lions in [2,3,4], where some impulse control problems are studied. They have become a powerful mathematical tool for modelling of various complicated equilibria which appeared in many areas such as mechanics [9, 16, 17], traffic [5], statistics [13], biology [10], economics and finance [18, 23]. For more analytical results of quasi-variational inequalities, we refer the reader to, for examples, [6, 19].
It is well-known that the problem (1.1) can be written as the following fixed point problem
where \(\lambda \) is a positive real number and \(P_C[\cdot ]\) is the projection operator on a closed convex set C in \(\mathbb {R}^n\). Thus, we can get sufficient conditions for the existence of solutions of a quasi-variational inequality by applying a suitable fixed point theorem. Some existence results in finite dimensional spaces can be found in [7]. Applying the Brouwer fixed point theorem, the authors in [7] proved the following existence result.
Theorem 1.1
[7] Suppose that there exists a nonempty compact convex set \(C\subset \mathrm {dom} K\) such that
-
(H1)
\(K(C)\subset C\),
-
(H2)
F is continuous on C,
-
(H3)
K is continuous on C.
Then the quasi-variational inequality (1.1) has at least one solution.
However, in some applications, e.g., the discretized Coulomb friction model (see, e.g., [16]), there is no compact set C for which condition (H1) is satisfied. In [16], to obtain the existence of solutions of the quasi-variational inequality (1.1), the authors replaced the compactness of C by a weaker condition, namely the closedness and required, in addition, that the function F is strongly monotone. More precisely, by using the Brouwer fixed point theorem, the authors in [16] proved the following existence result.
Theorem 1.2
[16] Let F be continuously differentiable on \(\mathbb {R}^n\) and be strongly monotone on \(\mathbb {R}^n\), i.e., there exists \(\alpha >0\) such that
Here, \(||\cdot ||\) denotes the norm induced by the inner product \(\langle \cdot ,\cdot \rangle \) in \(\mathbb {R}^n\). Suppose that there exists a nonempty convex closed set \(C\subset \mathrm {dom} K\) such that
-
(i)
\(K(C)\subset C\),
-
(ii)
\(\displaystyle \bigcap _{x\in C} K(x) \ne \emptyset ,\)
-
(iii)
for each \(y\in \mathbb {R}^n\), the mapping \(x\mapsto P_{K(x)}[y]\) is continuous on an open set containing C.
Then the quasi-variational inequality (1.1) has at least one solution.
For other existence results of of solutions of the quasi-variational inequality concerning strong monotonicity, we refer the reader to [22] and [20]. Both results need some contraction property of the projection operator on the set \(K(\cdot )\), i.e., they require the existence of a positive constant \(\ell <1\) such that
An example for \(K(\cdot )\) satisfying (1.4) is a moving set where \(K(x) = c(x)+K\) with K being a nonempty closed convex subset of \(\mathbb {R}^n\) and \(c:\mathbb {R}^n\rightarrow \mathbb {R}^n\) a Lipschitz continuous mapping with the same modulus \(\ell \). In [21], the authors extended the existence results in [20, 22] to the case when F is strongly pseudomonotone on \(\mathbb {R}^n\), i.e., there exists a modulus \(\gamma >0\) such that for all \(x,y\in \mathbb {R}^n\)
Since strongly monotone mappings are also strongly pseudomonotone, but not vice versa, there are examples that we can apply the existence result in [21] but we cannot apply any existence results in [16, 20, 22]. Note that, in general setting, (1.4) is hard to prove and it is not often satisfied. The following example shows that there are situations in which we cannot apply above mentioned results.
Example 1.1
We consider the quasi-variational inequality problem (1.1) with \(F: \mathbb {R}^2\rightarrow \mathbb {R}^2\) defined by:
and \(K:\mathbb {R}^2\rightrightarrows \mathbb {R}^2\)defined by:
Now, assume that \(x^* = (x_1^*,x_2^*) \in \mathbb {R}^2\) is a solution of the quasi-variational inequality (1.1). Then, \((x_1^*,x_2^*) \in [-2|x^*_1|,2|x^*_1|] \times [0,\infty )\) and
for all \((y_1,y_2) \in [-2|x^*_1|,2|x^*_1|] \times [0,\infty )\) . This is equivalent to \((x_1^*,x_2^*) \in [-2|x^*_1|,2|x^*_1|] \times [0,\infty )\) and
This implies that \(x_1^*=x_2^* = 0\). Thus, (0, 0) is a solution of the quasi-variational inequality (1.1).
We cannot apply Theorem 1.1 to this example since there is no compact set \(C\subset \mathbb {R}^2\) such that \(K(C)\subset C\). We also cannot apply Theorem 1.2 to this example since the mapping F is not strongly monotone. Indeed, taking \(x = (1,0)\) and \(y = (2,0)\), we have
One can see that F is strongly pseudomonotone on \(\mathbb {R}^n\) with modulus \(\displaystyle \frac{1}{6}\). Indeed, for \(x=(x_1,x_2), y=(y_1,y_2) \in \mathbb {R}^2\) with \(\langle F(x),y-x\rangle \ge 0\), i.e., \(\langle (x_1,2x_2),(y_1-x_1,2y_2-2x_2\rangle \ge 0\), we have
Moreover,
That is, the projection map is not contractive. So, we cannot apply [21, Theorem 3.1] to this example.
Our aim is to prove an existence result for strongly pseudomonotone quasi-variational inequalities without requiring (1.4). For this aim, we will extend Theorem 1.2 by replacing the strong monotonicity of F by the strong pseudomonotonicity. Our result bases on a new result about continuity property of solutions to parametric variational inequalities.
To conclude this section, we recall the following results which will be used in the sequel.
Lemma 1.1
(see, e.g., [15]) Let \(\varOmega \) be a nonempty closed convex subset of \(\mathbb {R}^n\). Then for a given \(z\in \mathbb {R}^n\), \(u\in \varOmega \) satisfies \(u = P_{\varOmega }[z]\) if and only if
Theorem 1.3
[15, Theorem 4.2, page 13] Let \(K\subset \mathbb {R}^n\) be closed and convex and \(F:K\rightarrow \mathbb {R}^n\) be continuous. A necessary and sufficient condition that there exists a solution to VIP (1.2) is that there exists an \(R>0\) such that \(||x_R||< R\), where \(x_R\in K\cap \overline{B}(\mathbf{0} ,R)\) satisfying
Here, \(\overline{B}(\mathbf{0} ,R)\) denotes the closed ball of radius R and center \(\mathbf{0} \), the zero vector of \(\mathbb {R}^n\).
2 Main results
In this section, by applying the Brouwer fixed point theorem, we derive an existence result for quasi-variational inequality (1.1) with strongly pseudomonotone mapping F. With this aim, we first prove a new result concerning continuity property of solution map to a parametric variational inequality.
2.1 Parametric variational inequalities
Let \(M\subset \mathbb {R}^m\) and \(\varLambda \subset \mathbb {R}^d\) be two sets of parameters. Let \(f:\mathbb {R}^n\times \mathbb {R}^m \rightarrow \mathbb {R}^n\) be a vector-valued function and \(K: \varLambda \rightrightarrows \mathbb {R}^n\) be a set-valued mapping. We consider the following parametric variational inequality problem with parameters \((\mu , \lambda )\in M\times \varLambda \):
Find \(x\in K(\lambda )\) such that
Throughout this subsection, we always require the following assumptions:
-
(A1)
\(K(\lambda )\) is nonempty, closed and convex for each \(\lambda \in \varLambda \).
-
(A2)
There exists \(\ell >0\) such that, for all \(x_1,x_2\in K(\varLambda ) \) and \(\mu _1,\mu _2\in M\),
$$\begin{aligned} ||f(x_1,\mu _1) - f(x_2,\mu _2)|| \le \ell (||x_1-x_2|| + ||\mu _1-\mu _2||). \end{aligned}$$(2.2) -
(A3)
There exists \(\alpha >0\) such that, for all \(\mu \in M\) and for all \(x,y\in K(\varLambda )\),
$$\begin{aligned} \langle f(x,\mu ),y-x\rangle \ge 0\quad \Rightarrow \quad \langle f(y,\mu ),y-x\rangle \ge \alpha ||x-y||^2. \end{aligned}$$
Under our assumptions, for each \((\lambda ,\mu ) \in M\times \varLambda \), the parametric variational inequality problem (2.1) has a unique solution (see [14, Theorem 2.1]). Hence, we can define a mapping \(u: M\times \varLambda \rightarrow \mathbb {R}^n\) assigning to each \((\lambda ,\mu ) \in M\times \varLambda \) the corresponding unique solution of (2.1).
The main result of this subsection is stated as follows.
Theorem 2.1
Let \((\lambda _0,\mu _0) \in M\times \varLambda \) and \(x_0 = u(\lambda _0,\mu _0)\). If the mapping \(\lambda \mapsto P_{K(\lambda )}[x_0-\rho f(x_0,\mu _0)]\) is continuous at \(\lambda _0\) for some \(\rho >1/\alpha \), then u is continuous at \((\lambda _0,\mu _0)\).
Proof
Fix \(\rho >1/\alpha \) and let \((\lambda ,\mu )\in M\times \varLambda \). We have
and
Set
By Lemma 1.1, one has
It follows that
Since \(y\in K(\lambda )\), we have
Thus, by (A3), one has
Using (A2), we have from (2.3) and (2.4) that
Here, we obtain the last inequality by using the inequality \(-2||a-b||^2 \le -||a||^2 + 2||b||^2\). Then,
Thus,
Since \(\rho \alpha >1\), we have \(2\rho ^2\ell ^2+2\rho \alpha -1 >\max \{1,2\rho ^2\ell ^2\}\). It follows from (2.5) that
Hence,
Since \(\lambda \mapsto P_{K(\lambda )}[x_0-\rho f(x_0,\mu _0)]\) is continuous at \(\lambda _0\), it follows from (2.6) that \(u(\lambda ,\mu )\) is continuous at \((\lambda _0,\mu _0)\). \(\square \)
Remark 2.1
It follows from (2.6) that properties of the solution mapping u depend on the properties of the mapping \(\lambda \mapsto P_{K(\lambda )}[x_0-\rho f(x_0,\mu _0)]\). In fact, if the mapping \(\lambda \mapsto P_{K(\lambda )}[x_0-\rho f(x_0,\mu _0)]\) is Hölder continuous (respectively, Lipschitz continuous), then the solution mapping u is also Hölder continuous (respectively, Lipschitz continuous). Examples of set-valued mappings K in which \(\lambda \mapsto P_{K(\lambda )}[x_0-\rho f(x_0,\mu _0)]\) is Hölder continuous or Lipschitz continuous can be seen, for instance, in [25, 26]. For examples of set-valued mappings K in which \(\lambda \mapsto P_{K(\lambda )}[x_0-\rho f(x_0,\mu _0)]\) is continuous, we refer the reader, e.g., to [16]. For more results concerning regularity property of solution maps to parametric variational inequalities, we refer the reader to, e.g., [1, 8, 12, 24,25,26].
The next result concerns the boundedness of the images of the solution mapping u.
Proposition 2.1
Assume that there exists a nonempty set \(\varOmega \subset \mathrm {dom}K\cap \varLambda \) such that
-
(A4)
\(\qquad \qquad \qquad \qquad \displaystyle \bigcap _{\lambda \in \varOmega }K(\lambda ) \ne \emptyset .\)
Then there exists \(R>0\) such that \(||u(\mu ,\lambda )|| <R\) for all \(\lambda \in \varOmega \) and \(\mu \in M\).
Proof
Let \(\mu _0\in M\) and \(\lambda _0\in \varOmega \). Let \(v_0\in \bigcap _{\lambda \in \varOmega }K(\lambda )\) and fix \(H>f(v_0,\mu _0)\). Set
Let \(z^*\) be the unique solution of the variational inequality: Find \(z\in K(\lambda _0)\cap \overline{B}(\mathbf{0} ,R)\) such that
Assume that \(||z^*|| \ge R\). Since \(v_0\in K(\lambda _0)\) and
\(v_0\in K(\lambda _0)\cap \overline{B}(\mathbf{0} ,R)\). Hence, we have
By (A3), one has
This implies that
We have
This is a contradiction. Thus, the solution \(z^*\) of (2.7) satisfies \(||z^*|| <R\) for some \(R>0\). It follows from Theorem 1.3 that \(z^*\) is also a solution of the the parametric variational inequality problem (2.1) with the pair of parameters \((\mu _0,\lambda _0)\). Therefore, \(||u(\mu _0,\lambda _0)|| <R\). This ends the proof. \(\square \)
2.2 Existence result
We are now in a position to state the main result of this paper.
Theorem 2.2
Let \(F: \mathbb {R}^n\rightarrow \mathbb {R}^n\) be a mapping and let \(K:\mathbb {R}^n\rightrightarrows \mathbb {R}^n\) be a set-valued mapping with closed convex values. Suppose that there exists a nonempty convex closed set \(C\subset \mathrm {dom} K\) such that
-
(i)
\(K(C)\subset C\),
-
(ii)
F is Lipschitz continuous on C and f is strongly pseudomonotone on C,
-
(iii)
\(\displaystyle \bigcap _{x\in C} K(x) \ne \emptyset ,\)
-
(iv)
for each \(y\in \mathbb {R}^n\), the mapping \(x\mapsto P_{K(x)}[y]\) is continuous on an open set containing C.
Then the quasi-variational inequality (1.1) has at least one solution.
Proof
We define a mapping \(S:\text{ dom }K\rightarrow \mathbb {R}^n\) by assigning to each \(y\in \mathrm {dom} K\) the corresponding unique solution of the following parameter variational inequality: Find \(z\in K(y)\) such that
By Theorem 2.1, S is continuous on C. Moreover, by Proposition 2.1, there exists \(R>0\) such that \(||S(y)|| <R\) for all \(y\in C\). That is, S maps \(C\cap \{x\in \mathbb {R}^n: ||x||\le R\}\) into itself. Hence, by the Brouwer fixed point theorem, S has a fixed point in \(C\cap \{x\in \mathbb {R}^n: ||x||\le R\}\), which is a solution of (1.1). This ends the proof. \(\square \)
Example 2.1
We consider F and K as in Example 1.1. As shown in Example 1.1, the mapping F is strongly pseudomonotone. One can also see that \(C = \{0\}\times [0,\infty )\) is closed convex such that condition (i) and (iii) of Theorem 2.2 are satisfied. It is easy to see that F is Lipschitz continuous on C. For \(x = (x_1,x_2)\in \mathbb {R}^2\) and \(y=(y_1,y_2)\in \mathbb {R}^2\), one can compute that
Then the map \(x\mapsto P_{K(x)}[y]\) is continuous for each \(y\in \mathbb {R}^2\) and so the condition (iv) of Theorem 2.2 is satisfied. Therefore, we can apply Theorem 2.2 to this example.
Example 2.2
(see, e.g. [16]) Let m be an even integer and consider \(K: \mathbb {R}^m \rightarrow \mathbb {R}^m\) which is defined by: for \(x = (x_1,\cdots ,x_m)\in \mathbb {R}^m\),
where \(\alpha >0\) is a constant. This type of set-valued mappings appears in the discretized Coulomb friction model (see, [11]).
If we set \(C = \{x = (x_1,\ldots ,x_m)\in \mathbb {R}^m: x_{2i-1} \le 0, i=1,2,\ldots ,m/2\}\), then conditions (i), (iii) and (iv) of Theorem 2.2 are satisfied.
To conclude this paper, we present a general class of strongly pseudomonotone mappings which are not strongly monotone in general.
Example 2.3
(see, e.g., [14]) Let \(F:\mathbb {R}^n\rightarrow \mathbb {R}^n\) be defined by
where \(Q:\mathbb {R}^n \rightarrow \mathbb {R}^n\) is a linear mapping satisfying
for some constant \(\gamma >0\), \(h: \mathbb {R}^n\rightarrow [a,\infty )\) for some constant \(\alpha >0\) and \(q\in \mathbb {R}^n\).
Let \(x,y\in \mathbb {R}^n\) be such that \(\langle F(x),y-x\rangle \ge 0\), i.e., \(\langle h(x)(Qx+q),y-x\rangle \ge 0\). This implies
Thus,
Hence, F is strongly pseudomonotone with modulus \(\alpha \gamma \). Note that the mapping F considered in Example 1.1 is a particular case of the form (2.8). So, in general, a mapping of the form (2.8) is not strongly monotone.
References
Alt, W., Kolumbán, I.: An implicit-function theorem for a class of monotone generalized equations. Kybernetika 29, 210–221 (1993)
Bensoussan, A., Goursat, M., Lions, J.L.: Contrôle impulsionnel et inéquations quasi-variationnelles. C. R. Acad. Sci. Paris Sér A 276, 1284–1973 (1973)
Bensoussan, A., Lions, J..L.: Nouvelle formulation de problèmes de contrôle impulsionnel et applications. C. R. Acad. Sci. Paris Sér. A 276, 1189–1192 (1973)
Bensoussan, A., Lions, J.L.: Nouvelles méthodes en contrôle impulsionnel. Appl. Math. Optim. 1, 289–312 (1975)
Bliemer, M., Bovy, P.: Quasi-variational inequality formulation of the multiclass dynamic traffic assignment problem. Transp. Res. B Methodol. 37, 501–519 (2003)
Baiocchi, C., Capelo, A.: Variational and Quasivariational Inequalities: Applications to Free Boundary Problems. Wiley, New York (1984)
Chan, D., Pang, J.S.: The generalized quasi-variational inequality problem. Math. Oper. Res. 7, 211–222 (1982)
Dafermos, S.: Sensitivity analysis in variational inequalities. Math. Oper. Res. 13, 421–434 (1988)
Haslinger, J., Panagiotopoulos, P..D.: The reciprocal variational approach to the Signorini problem with friction. Approximation results. Proc. R. Soc. Edinburgh Sect. A Math. 98, 365–383 (1994)
Hammerstein, P., Selten, R.: Game theory and evolutionary biology. In: Aumann, R.J., Hart, S. (eds.) Handbook of Game Theory with Economic Applications, vol. 2, pp. 929–993. North-Holland, Amsterdam (1994)
Hlavác̆ek, I., Haslinger, J., Nec̆as, J., Lovís̆ek, J.: Solution of variational inequalities in mechanics. Springer, New York (1988)
Kassay, G., Kolumbán, J.: Multivalued parametric variational inequalities with \(\alpha \)-pseudomonotone maps. J. Optim. Theory Appl. 107, 35–50 (2000)
Kharroubi, I., Ma, J., Pham, H., Zhang, J.: Backward SDEs with constrained jumps and quasi-variational inequalities. Ann. Probab. 38, 794–840 (2010)
Kim, D.S., Vuong, P.T., Khanh, P.D.: Qualitative properties of strongly pseudomonotone variational inequalities. Optim. Lett. 10, 1669–1679 (2016)
Kinderlehrer, D., Stampacchia, G.: An Introduction to Variational Inequalities and Their Applications. Academic Press, New York (1980)
Kocvara, M., Outrata, J.: On a class of quasi-variational inequalities. Optim. Methods Softw. 5, 275–295 (1995)
Kravchuk, A.S., Neittaanmaki, P.J.: Variational and Quasi-Variational Inequalities in Mechanics. Springer, Dordrecht (2007)
Lahmdani, A., Chadli, O., Yao, J.C.: Existence of solutions for noncoercive hemivariational inequalities by an equilibrium approach under pseudomonotone perturbation. J. Optim. Theory Appl. 160(2014), 49–66 (2014)
Mosco, U.: Implicit variational problems and quasi variational inequalities. In: Gossez, J., Lami Dozo, E., Mawhin, J., Waelbroeck, L. (eds.) Nonlinear Operators and the Calculus of Variations. Lecture Notes in Mathematics, vol. 543, pp. 83–156. Springer, Berlin (1976)
Nesterov, Y., Scrimali, L.: Solving strongly monotone variational and quasi-variational inequalities. Discrete Contin. Dyn. Syst. 31, 1383–1396 (2011)
Nguyen, L.V., Qin, X.: Some results on strongly pseudomonotone quasi-variational inequalities. Set Valued Var. Anal. 28, 239–257 (2020)
Noor, M.A., Oettli, W.: On general nonlinear complementarity problems and quasiequilibria. Le Matematiche XLIX 49, 313–331 (1994)
Wei, J.Y., Smeers, Y.: Spatial oligopolistic electricity models with Cournot generators and regulated transmission prices. Oper. Res. 47, 102–112 (1999)
Yao, J.C.: Multivalued variational inequalities with K-pseudomonotone operators. J. Optim. Theory Appl. 83, 391–403 (1994)
Yen, N.D.: Lipschitz continuity of solutions of variational inequalities with a parametric polyhedral constraint. Math. Oper. Res. 20, 695–708 (1995)
Yen, N.D.: Holder continuity of solutions to a parametric variational inequality. Appl. Math. Optim. 31, 245–255 (1995)
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares that he has no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Van Nguyen, L. An existence result for strongly pseudomonotone quasi-variational inequalities. Ricerche mat 72, 803–813 (2023). https://doi.org/10.1007/s11587-021-00588-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11587-021-00588-y