Abstract
In this paper, inspired by various stochastic linear-quadratic (LQ, for short) problems, we develop the method of continuation to study nonlinear forward–backward stochastic differential equations (FBSDEs, for short) in a kind of domination-monotonicity frameworks. The coupling of FBSDEs is in a general form, i.e., it not only appears in integral terms and terminal terms, but also in initial terms. By virtue of introducing various matrices, matrix-valued random variables and matrix-valued stochastic processes, we present the domination-monotonicity framework carefully and rigorously. A unique solvability result and a pair of estimates for coupled FBSDEs are obtained (see Theorem 3.5 in the case of simple domination-monotonicity conditions and Theorem 5.2 in the case of multi-level self-similar domination-monotonicity structures). As applications of theoretical results, the related stochastic Hamiltonian systems of several LQ problems are discussed.
Similar content being viewed by others
References
Antonelli, F.: Backward-forward stochastic differential equations. Ann. Appl. Probab. 3(3), 777–793 (1993)
Bensoussan, A., Chen, S., Sethi, S.P.: The maximum principle for global solutions of stochastic Stackelberg differential games. SIAM J. Control Optim. 53(4), 1956–1981 (2015)
Briand, P., Coquet, F., Hu, Y., Mémin, J., Peng, S.: A converse comparison theorem for BSDEs and related properties of \(g\)-expectation. Electron. Commun. Probab. 5, 101–117 (2000)
Delarue, F.: On the existence and uniqueness of solutions to FBSDEs in a non-degenerate case. Stoch. Process. Appl. 99(2), 209–286 (2002)
Hamadène, S.: Backward-forward SDE’s and stochastic differential games. Stoch. Process. Appl. 77(1), 1–15 (1998)
Hu, Y., Peng, S.: Solution of forward-backward stochastic differential equations. Probab. Theory Related Fields 103(2), 273–283 (1995)
Li, N., Yu, Z.: Forward-backward stochastic differential equations and linear-quadratic generalized Stackelberg games. SIAM J. Control Optim. 56(6), 4148–4180 (2018)
Ma, J., Protter, P., Yong, J.: Solving forward-backward stochastic differential equations explicitly: a four step scheme. Probab. Theory Related Fields 98(3), 339–359 (1994)
Ma, J., Wu, Z., Zhang, D., Zhang, J.: On well-posedness of forward-backward SDEs: a unified approach. Ann. Appl. Probab. 25(4), 2168–2214 (2015)
Ma, J., Yong, J.: Forward-backward stochastic differential equations and their applications. Lecture Notes in Math, vol. 1702. Springer, Berlin (1999)
Pardoux, É., Peng, S.: Adapted solution of a backward stochastic differential equation. Syst. Control Lett. 14(1), 55–61 (1990)
Peng, S.: Problem of eigenvalues of stochastic Hamiltonian systems with boundary conditions. Stoch. Process. Appl. 88(2), 259–290 (2000)
Peng, S., Wu, Z.: Fully coupled forward-backward stochastic differential equations and applications to optimal control. SIAM J. Control Optim. 37(3), 825–843 (1999)
Shi, J., Wang, G., Xiong, J.: Leader-follower stochastic differential game with asymmetric information and applications. Automatica 63, 60–73 (2016)
von Stackelberg, H.: Marktform und Gleichgewicht, Springer, Wien New York, 1934 (in German). Market Structure and Equilibrium (2011). (in English)
Yong, J.: Finding adapted solutions of forward-backward stochastic differential equations: method of continuation. Probab. Theory Related Fields 107(4), 537–572 (1997)
Yong, J.: A leader-follower stochastic linear quadratic differential game. SIAM J. Control Optim. 41(4), 1015–1041 (2002)
Yong, J.: Forward-backward stochastic differential equations with mixed initial-terminal conditions. Trans. Am. Math. Soc. 362(2), 1047–1096 (2010)
Yong, J., Zhou, X.: Stochastic Controls: Hamiltonian Systems and HJB Equations. Springer, New York (1999)
Yu, Z.: An optimal feedback control-strategy pair for zero-sum linear-quadratic stochastic differential game: the Riccati equation approach. SIAM J. Control Optim. 53(4), 2141–2167 (2015)
Zhang, J.: The wellposedness of FBSDEs. Discrete Contin. Dyn. Syst. Ser. B 6(4), 927–940 (2006)
Zhang, J.: Backward stochastic differential equations: from linear to fully nonlinear theory, Probability Theory and Stochastic Modelling, 86. Springer, New York (2017)
Zheng, Y., Shi, J.: A Stackelberg game of backward stochastic differential equations with applications. Dyn. Games Appl. 10(4), 968–992 (2020)
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work is supported by the National Natural Science Foundation of China (11871310).
Appendix A: A Sketch of the Proof of Theorem 5.2
Appendix A: A Sketch of the Proof of Theorem 5.2
As we stated at the end of Sect. 5, the induction method and the continuation method will be combined together to give the proof of Theorem 5.2. Clearly, Theorem 3.5 means that Theorem 5.2 holds true when \(p=1\). Next, for any \(p\ge 2\), we assume that Theorem 5.2 holds for \(p-1\), then we need prove the conclusions also hold for p.
We notice that there are two cases in the monotonicity conditions at each qth level (\(q=1,2,\dots ,p\)). Overall, we have \(2^p\) situations. In the following, we only consider one of them: All of the monotonicity conditions take Cases 1. The other \(2^p-1\) situations can be treated in the same way or in the viewpoint of transformation (see Remark 3.3).
As the start of the method of continuation, now we define a set of coefficients \((\Psi ^0,\Phi ^0,\Gamma ^0)\). In detail, let
and
Moreover, we denote
and
We notice that there are \(2^{p-q}\) non-zero submatrices involved in the above defined \({\mathbb {M}}_q\), \({\mathbb {G}}_q\) and \(\mathbb L_q(\cdot )\) respectively for any \(q=1,2,\dots ,p\). Furthermore, we define
for any \((\omega ,s,\theta ) \in \Omega \times [0,T] \times \mathbb R^{2^{p-1}n(2+d)}\). For the sake of intuition, let us give the form of the matrix \(\sum _{q=1}^4 {\mathbb {M}}_q\) when \(p=4\) as follows:
where the blanks mean being filled by zeros. The matrix-valued random variable \(\sum _{q=1}^4 {\mathbb {G}}_q\) and the matrix-valued stochastic process \(\sum _{q=1}^4 {\mathbb {L}}_q(\cdot )\) are in similar forms. Similar to [7, Lemma 2.7], we can directly verify that the coefficients \((\Psi ^0, \Phi ^0, \Gamma ^0)\) defined by (A.6) satisfy Assumptions (H\(^p_0\))—(H\(^p_p\)) with all Cases 1.
Now, for any \((\xi ,\eta ,\rho (\cdot )) \in {\mathcal {H}}[0,T;p]\) with \(\rho (\cdot ) = (\varphi (\cdot )^\top ,\psi (\cdot )^\top ,\gamma (\cdot )^\top )^\top \), we introduce a family of FBSDEs parameterized by \(\alpha \in [0,1]\) on the interval [0, T] as follows:
where the coefficients \((\Psi ^\alpha , \Phi ^\alpha , \Gamma ^\alpha )\) are defined as a convex combination of \((\Psi ^0,\Phi ^0,\Gamma ^0)\) (see (A.6)) and the original set of coefficients \((\Psi ,\Phi ,\Gamma )\):
Since both \((\Psi ^0,\Phi ^0,\Gamma ^0)\) and \((\Psi ,\Phi ,\Gamma )\) satisfy Assumptions (H\(^p_0\))—(H\(^p_p\)) with all Cases 1, then it is clear that \((\Psi ^\alpha , \Phi ^\alpha , \Gamma ^\alpha )\) (and then \((\Psi ^\alpha +\xi , \Phi ^\alpha +\eta , \Gamma ^\alpha +\rho )\)) also satisfy Assumptions (H\(^p_0\))—(H\(^p_p\)) with all Cases 1.
When \(\alpha =0\), due to the delicate setting (A.6) of \((\Psi ^0,\Phi ^0,\Gamma ^0)\), FBSDE (A.7) is actually in a decoupled form with any given \((\xi ,\eta ,\rho (\cdot )) \in {\mathcal {H}}[0,T;p]\), therefore it is unique solvable. The detailed verification is very similar to [7, Lemma 3.3] and is omitted. Another fact is that, when \(\alpha =1\) and \((\xi ,\eta ,\rho (\cdot ))\) vanish, FBSDE (A.7) coincides with FBSDE (5.1). From the known \(\alpha =0\) case to the unknown \(\alpha =1\) case, we need to prove a continuity lemma similar to Lemma 3.7. In the establishing of the continuity lemma, the following a priori estimate will play a key role.
Lemma A.1
Let \(p\in {\mathbb {N}}\) and a set of coefficients \((\Psi ,\Phi ,\Gamma )\) are given to satisfy Assumptions (H\(^p_0\))—(H\(^p_p\)) with all Cases 1. Let \(\alpha \in [0,1]\) and \((\xi ,\eta ,\rho (\cdot ))\), \(({{\bar{\xi }}}, {{\bar{\eta }}},{{\bar{\rho }}}(\cdot )) \in {\mathcal {H}}[0,T;p]\). Suppose that \(\theta (\cdot )\), \({{\bar{\theta }}}(\cdot ) \in (M^2_{\mathbb F}(0,T;{\mathbb {R}}^{n(2+d)}))^{2^{p-1}}\) satisfy FBSDE (A.7) with the coefficients \((\Psi ^\alpha +\xi ,\Phi ^\alpha +\eta , \Gamma ^\alpha +\rho )\) and \((\Psi ^\alpha +{{\bar{\xi }}},\Phi ^\alpha +{{\bar{\eta }}}, \Gamma ^\alpha +{{\bar{\rho }}})\), respectively. Then the estimate (3.13) holds, where the constant K depending on p, T, the Lipschitz constants, \(\mu _1\), \(\nu _1\), \(\mu _q\), and the bounds of all \({{\widetilde{M}}}_1\), \({{\widetilde{G}}}_1\), \(\widetilde{A}_1(\cdot )\), \({{\widetilde{B}}}_1(\cdot )\), \({{\widetilde{C}}}_{1,j}(\cdot )\), \(M_q\), \(G_q\), \(L_q(\cdot )\), \(j=1,2,\dots ,d\), \(q=2,3,\dots ,p\).
Proof
We split the index set \(\{1,2,3,\dots ,2^{p-1}\}\) into two groups. Similar to Sect. 5, we denote
Firstly, due to the self-similar property of the domination-monotonicity structure, the following holds true: Being frozen \(\theta _{[\kappa _3, \kappa _4]}\), both \((\Psi ^\alpha +\xi ,\Phi ^\alpha +\eta , \Gamma ^\alpha +\rho )_{[\kappa _1,\kappa _2]}\) and \((\Psi ^\alpha +{{\bar{\xi }}},\Phi ^\alpha +{{\bar{\eta }}}, \Gamma ^\alpha +{{\bar{\rho }}})_{[\kappa _1,\kappa _2]}\) satisfy Assumptions (H\(^{p-1}_0\))—(H\(^{p-1}_{p-1}\)) with all Cases 1. This fact, with the help of Theorem 5.2 for \((p-1)\), leads to
where
and
By the definition (A.8) of \((\Psi ^\alpha ,\Phi ^\alpha ,\Gamma ^\alpha )\) and the domination conditions (see Assumption (H\(^p_p\))-(i)), we calculate
Similarly,
Therefore, we have
Secondly, being frozen \(\theta _{[\kappa _1,\kappa _2]}\), it is clear that, both \((\Psi ^\alpha +\xi ,\Phi ^\alpha +\eta , \Gamma ^\alpha +\rho )_{[\kappa _3,\kappa _4]}\) and \((\Psi ^\alpha +{{\bar{\xi }}},\Phi ^\alpha +{{\bar{\eta }}}, \Gamma ^\alpha +{{\bar{\rho }}})_{[\kappa _3,\kappa _4]}\) satisfy Assumptions (H\(^{p-1}_0\))—(H\(^{p-1}_{p-2}\)) with all Cases 1 and (H\(^{p-1}_{p-1}\)) with Case 2. Theorem 5.2 for \((p-1)\) works once again to yields
where
and
By the definition (A.8) of \((\Psi ^\alpha ,\Phi ^\alpha ,\Gamma ^\alpha )\) and the Lipschitz conditions (see Assumption (H\(^p_0\))-(ii)), we calculate
Similarly,
Then,
By combining (A.10) and (A.11), we have
where \({{\widehat{J}}}\) is defined by (3.14).
Thirdly, applying Itô’s formula to \(\langle {{\widehat{x}}}(\cdot ),\ \Lambda {{\widehat{y}}}(\cdot ) \rangle = \sum _{j=\kappa _1}^{\kappa _4} \langle {{\widehat{x}}}_j(\cdot ),\ \widehat{y}_{\kappa _1+\kappa _4-j}(\cdot ) \rangle \) yields
By the monotonicity conditions of \((\Psi ^\alpha ,\Phi ^\alpha ,\Gamma ^\alpha )\) (see(H\(^p_p\))-(iii)-Case 1), we have
Similar to (3.18), by Young’s inequality, for any \(\varepsilon >0\), we have
By taking \(\varepsilon =\mu _p/(4K_3)\) where \(K_3\) is the constant appearing in (A.12), we obtain the a priori estimate (3.13) from (A.12) and (A.13). \(\square \)
Next, with the help of Lemma A.1, we establish the following continuation lemma.
Lemma A.2
Let \(p\in {\mathbb {N}}\) and a set of coefficients \((\Psi ,\Phi ,\Gamma )\) are given to satisfy Assumptions (H\(^p_0\))—(H\(^p_p\)) with all Cases 1. Then there exists an absolute constant \(\delta _0>0\) such that, if for some \(\alpha _0\in [0,1)\), FBSDE (A.7) is unique solvable in the space \((M^2_{\mathbb F}(0,T;{\mathbb {R}}^{n(2+d)}))^{2^{p-1}}\) for any \((\xi ,\eta ,\rho (\cdot )) \in {\mathcal {H}}[0,T;p]\), then the same result is also true for \(\alpha =\alpha _0+\delta \) with \(\delta \in [0,\delta _0]\) and \(\alpha \le 1\).
Proof
Let \(\delta _0>0\) be determined below. Let \(\delta \in [0,\delta _0]\). Similar to the proof Lemma 3.7, we shall define a mapping through FBSDEs with \(\alpha =\alpha _0\). In detail, for any \(\theta (\cdot ) \in (M^2_{{\mathbb {F}}}(0,T;\mathbb R^{n(2+d)}))^{2^{p-1}}\), we consider an FBSDE with unknown \(\Theta (\cdot ) = (X(\cdot )^\top ,Y(\cdot )^\top ,Z(\cdot )^\top )^\top \) as follows:
where
By our assumption, FBSDE (A.14) is unique solvable in \((M^2_{{\mathbb {F}}}(0,T;{\mathbb {R}}^{n(2+d)}))^{2^{p-1}}\). Therefore, we have established a mapping \(\mathscr {T}_{\alpha _0+\delta }\) by
By carefully observing the definition (A.8) of the coefficients \((\Psi ^{\alpha _0+\delta },\Phi ^{\alpha _0+\delta },\Gamma ^{\alpha _0+\delta })\), it is clear that the fixed point of the mapping \(\mathscr {T}_{\alpha _0+\delta }\) is equivalent to the solution of FBSDE (A.7) with \(\alpha =\alpha _0+\delta \). Next, we shall prove that the mapping \({\mathscr {T}}_{\alpha _0+\delta }\) admits a unique fixed point by Banach’s Contraction Mapping Theorem.
For any given \(\theta (\cdot )\), \({{\bar{\theta }}}(\cdot ) \in (M^2_{\mathbb F}(0,T;{\mathbb {R}}^{n(2+d)}))^{2^{p-1}}\), denote \(\Theta (\cdot ) ={\mathscr {T}}_{\alpha _0+\delta }(\theta (\cdot ))\) and \({{\bar{\Theta }}}(\cdot ) ={\mathscr {T}}_{\alpha _0+\delta }({{\bar{\theta }}}(\cdot ))\). Moreover, we denote \({{\widehat{\theta }}}(\cdot ) = \theta (\cdot ) -{{\bar{\theta }}}(\cdot )\), \({{\widehat{\Theta }}}(\cdot ) = \Theta (\cdot ) -{{\bar{\Theta }}}(\cdot )\), and so on. We apply Lemma A.1 to get
Select \(\delta _0 =1/(2\sqrt{K_4})\). Then, for any \(\delta \in [0,\delta _0]\), the above inequality implies that \(\mathscr {T}_{\alpha _0+\delta }\) is a contraction. The proof is finished. \(\square \)
Similar to the proof of Theorem 3.5, with the help of Lemmas A.1 and A.2, we can easily prove Theorem 5.2 under Assumptions (H\(^p_0\))–(H\(^p_p\)) with all Cases 1.
Rights and permissions
About this article
Cite this article
Yu, Z. On Forward–Backward Stochastic Differential Equations in a Domination-Monotonicity Framework. Appl Math Optim 85, 5 (2022). https://doi.org/10.1007/s00245-022-09841-8
Accepted:
Published:
DOI: https://doi.org/10.1007/s00245-022-09841-8
Keywords
- Forward–backward stochastic differential equation
- Stochastic linear-quadratic problem
- Method of continuation
- Stochastic optimal control