Abstract
In this paper, we introduce a new structure named “progressive structure” to deal with stochastic control problem with jumps. One example is given to show the motivation of our new structure compared with the traditional one at the beginning. And then, we obtain the maximum principle for a forward stochastic control problem by virtue of a new variation method. The control is allowed to enter both diffusion and jump terms and the control domain is convex. In the end, we apply our theoretical results to another example to illustrate the efficiency of our method.
Similar content being viewed by others
References
He, S., Wang, J., Yan, J.: Semimartingale Theory and Stochastic Calculus. Routledge, Abingdon (1992)
Peng, S.: A general stochastic maximum principle for optimal control problems. SIAM J. Control Optim. 28(4), 966–979 (1990)
Peng, S.: Backward stochastic differential equations and applications to optimal control. Appl. Math. Optim. 27(2), 125–144 (1993)
Situ, R.: Theory of Stochastic Differential Equations with Jumps and Applications: Mathematical and Analytical Techniques with Applications to Engineering. Springer, New York (2006)
Shi, J., Wu, Z.: Maximum principle for forward-backward stochastic control system with random jumps and applications to finance. J. Syst. Sci. Complex. 23(2), 219–231 (2010)
Song, Y., Tang, S., Wu, Z.: The maximum principle for progressive optimal stochastic control problems with random jumps. SIAM J. Control Optim. 58(4), 2171–2187 (2020)
Tang, S., Li, X.: Necessary conditions for optimal control of stochastic systems with random jumps. SIAM J. Control Optim. 32(5), 1447–1475 (1994)
Wu, Z.: Maximum principle for optimal control problem of fully coupled forward-backward stochastic systems. J. Systems Sci. Math. Sci. 11, 249–259 (1998)
Yan, J.: Introduction to Martingales and Stochastic Integrals. Shanghai Sci. and Tech. Publ. House, Shanghai (1981)
Acknowledgements
This work was supported by the Natural Science Foundation of China (11831010, 61961160732), Shandong Provincial Natural Science Foundation (ZR2019ZD42) and the Taishan Scholars Climbing Program of Shandong (TSPD20210302).
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Nizar Touzi.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: The Existence and Uniqueness of Solution of SDE
We are given the following SDE
where \(x_0\in R^n\), \(b,a,\sigma ,c:\varOmega \times [0, T]\times Z\times R^n \rightarrow R^n\), n is the dimension of X. We introduce a Banach space
with norm \(\Vert X\Vert ^2=E\left[ \sup _{0\le t\le T}|X_t|^2\right] \).
Lemma A.1
Suppose that \(X_t\) is an adapted process with càdlàg paths such that
for some \(p>0\). Then, we have
where C is a constant only related to p.
Proof
Set \(A_t=\int _0^t |X_{s-}| N(ds,Z)\). Since \(N([0,t]\times Z)\) is a pure jump process, \(A_t\) is a pure jump process. Notice that the jump time of \(A_t\) is also a jump time of \(N([0,t]\times Z)\) and the jump size of \(N([0,t]\times Z)\) is always equal to 1, so we have
for any \(k\ge 1\). Since \(A_{\cdot -}\), \(X_{\cdot -}\) and \(I_{[0,T_k]}\) are predictable, we have
Since \(E\left[ A_{s\wedge T_k}^p\right] <kE\left[ \sup _{0\le t\le T}|X_t|^p\right] \), by Gronwall’s inequality, we have
Let k goes to infinity, by Fatou’s lemma, we have
\(\square \)
We have the following assumptions:
Assumption H1:
-
(i)
\(b,a,\sigma ,c\) are \({\mathscr {G}} \otimes {\mathscr {Z}}\otimes {\mathscr {B}}(R^n)/{\mathscr {B}}(R^n)\) measurable.
-
(ii)
\(b, a,\sigma , c\) are uniform Lipschitz continuous with respect to x.
-
(iii)
\(E\left( \int _0^T \left( \int _Z|b(\omega ,t,e, 0)|\lambda (de)\right) dt\right) ^2<\infty ,E\int _0^T \left( \int _Z|\sigma (\omega ,t,e, 0)|\lambda (de)\right) ^2dt<\infty \), \(E\left( \int _0^T\int _Z |a(\omega ,t,e,0)|N(dt, de)\right) ^2<\infty , E\int _0^T\int _Z |c(\omega ,t,e, 0)|^2N(dt, de)<\infty \).
Theorem A.1
Under Assumption H1, (24) has a unique solution in \(S^2[0, T]\).
Proof
Firstly, we show that there is a unique solution in small time duration. We construct a map from \(S^2[0, T]\) to \(S^2[0, T]\)
It is easy to show that the image of \({\mathscr {T}}\) is actually in \(S^2[0, T]\) by (ii) and (iii) in Assumption H1 and Lemma A.1. For any \(X, Y\in S^2[0, T]\),
We choose T small enough that \(C\left( T+T^2+e^{CT}T\right) <1\), then \({\mathscr {T}}\) is a contraction.
For arbitrary T, we can split T into finite small pieces, so that we get a unique solution on each piece and connect them together. \(\square \)
Now, we give the \(L^2\) estimate of the solution.
Theorem A.2
Suppose that we are given two SDEs:
where \(i=1,2\). Suppose the two equations satisfy Assumption H1, and \(X^1\) (resp. \(X^2\)) is the solution of the first (resp. second) equation, then we have the following estimate:
Proof
We first suppose that T is sufficiently small. Denote \({\mathscr {T}}^i\) the contraction mapping with respect to the ith equation. Set
then, by the argument in Theorem A.1, we have
Choosing T small enough such that \(L(T)<1\), we have
which is the estimate (26). For arbitrary T, we can split T into finite small pieces, and then we connect the estimate and get the result. \(\square \)
Appendix B: Proof that \(\llbracket T_{n}, U_n\rrbracket \) is Z-Progressive
Lemma B.1
For each \(n\ge 1\), \(\llbracket T_{n}, U_n\rrbracket \) is Z-progressive.
Proof
For any \(t\ge 0\) and \(U\in {\mathscr {Z}}\), we have
Since for any \(U\in {\mathscr {Z}}\), \(X_t:=N(\omega ,[0,t]\times U)\) is a progressive process, \(X_{T_1}=I_{\{U_1\in U\}}\) is \({\mathscr {F}}_{T_1}\) measurable. Inductively, we can show that for each \(n\ge 1\), \(U_n\) is \({\mathscr {F}}_{T_n}\) measurable.
For simplicity, we consider Z to be R (real number) here. For fixed n, define the following set:
For fixed e, since \(U_n\) is \({\mathscr {F}}_{T_n}\) measurable, \(A(e)= \{T_n(\omega )\le t\}\cap \{U_n(\omega )\le e\}\) is a progressive set. For fixed \((\omega , t)\), \(I_{A(\omega ,t)}(e)\) is a right continuous function on R. Therefore, A is Z-progressive. By the same way, we can show that
are also Z-progressive. Then, \(\llbracket T_{n}, U_n\rrbracket =A-(A_1\cup A_2)\) is Z-progressive. \(\square \)
Appendix C: Complement Proof of Lemma 4.1
In the proof of Lemma 4.1, the convergence of the term related to a is not proved. In order to prove that we need some lemmas. First, let us introduce some notations.
Suppose \((E,{\mathscr {E}},\mu )\) is a measure space such that \(\mu \) is a finite measure. \((F,{\mathscr {F}})\) is a measurable space. \(K:E\times {\mathscr {F}}\rightarrow R_+\) is a finite transition kernel from E to F such that \(\mu \times K\) is a finite measure on \((E\times F,{\mathscr {E}}\otimes {\mathscr {F}})\).
Lemma C.1
Let \(f_n,f: E\times F\rightarrow R\) be \({\mathscr {E}}\otimes {\mathscr {F}}\) measurable functions such that \(f_n\) converges to f in measure \(\mu \times K\). If there exists a \({\mathscr {E}}\otimes {\mathscr {F}}\) measurable function g such that \(|f_n|\le g\) and \(\int _{F}g(x,y)K(x,dy)<\infty \), \(\mu \)-almost surely, then \(\int _Ff_n(x,y)K(x,dy)\) converges to \(\int _Ff(x,y)K(x,dy)\) in measure \(\mu \) as n tends to infinity.
Proof
We set \(h_n(x):=\int _Ff_n(x,y)K(x,dy)\) and \(h(x):=\int _Ff(x,y)K(x,dy)\). In order to show that \(h_n\) converges to h in \(\mu \), since \(\mu \) is a finite measure, we show for every subsequence \(h_{n_k}\) of \(h_n\), there is a subsequence \(h_{n_{k_l}}\) of \(h_{n_k}\) such that \(h_{n_{k_l}}\) converge to h \(\mu \) almost surely. Suppose \(h_{n_k}\) is any subsequence of \(h_n\), as \(f_{n_k}\) converges to f in measure \(\mu \times K\), then there exists a subsequence \(f_{n_{k_l}}\) of \(f_{n_k}\) such that \(f_{n_{k_l}}\) converges to f \(\mu \times K\) almost surely. Suppose \(A\in {\mathscr {E}}\otimes {\mathscr {F}}\) is the almost sure set.
Now, define the section set of A, for any \(x\in E\),
It is obvious that \(A_x\in {\mathscr {F}}\). Since
we have for \(\mu \)-a.s. \(x\in E\), \(K(x,A_x)=K(x,F)\), which means that \(A_x\) is a \(K(x,\cdot )\) almost sure set. Let B be the \(\mu \) almost sure set of E such that \(K(x,A_x)=K(x,F)\) and let C be the \(\mu \) almost sure set of E such that \(\int _{F}g(x,y)K(x,dy)<\infty \). Set \(D=B\cap C\), then for any \(x\in D\), \(f_{n_{k_l}}(x,\cdot )\) converges to \(f(x,\cdot )\), \(K(x,\cdot )\)-a.s., and are bounded by an integrable function \(g(x,\cdot )\). By dominate convergence theorem, we have \(h_{n_{k_l}}\) converges to h, \(\mu \)-a.s. \(\square \)
Lemma C.2
Suppose that the conditions in Lemma C.1 are satisfied except for the integrability condition of g turning into the following condition:
then,
Proof
By (28), we have \(\int _{F}g(x,y)K(x,dy)<\infty \), \(\mu \)-a.s., so \(\int _Ff_n(x,y)K(x,dy)\) converges to \(\int _Ff(x,y)K(x,dy)\) in measure \(\mu \). Then, we get the result by applying dominate convergence theorem. \(\square \)
Now, in order to show the convergence of the term related to a in Lemma 4.1, we set \((E,{\mathscr {E}},\mu )=(\varOmega , {\mathscr {F}},P)\), \((F,{\mathscr {F}})=([0,T]\times Z\times [0,1], {\mathscr {B}}([0,T])\otimes {\mathscr {Z}}\otimes {\mathscr {B}}([0,1]))\), \(K(\omega ,A\times B\times C)=Leb(C)N(\omega ,A\times B)\), \(g=M\left( |{\hat{X}}_t|^2+|v_t-u_t|^2\right) \). Then, applying Lemma C.2, we can obtain the convergence.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Song, Y., Wu, Z. The Maximum Principle for Stochastic Control Problem with Jumps in Progressive Structure. J Optim Theory Appl 199, 415–438 (2023). https://doi.org/10.1007/s10957-023-02302-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10957-023-02302-4