Constrained stochastic linear quadratic control under regime switching with controlled jump size
Xiaomin Shi, Zuo Quan Xu · 2024
Read Constrained stochastic linear quadratic control under regime switching with controlled jump size by Xiaomin Shi, Zuo Quan Xu (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-19100.txt.
Introduction
Xiaomin Shi and Zuo Quan Xu solve, in a 2024 math paper, a constrained control problem for systems that switch regimes and jump. Such systems model markets or machines whose rules change by state, with sudden shocks on top. Two controls act: a steady one in the smooth part, and an odd one that sets the jump size itself.
Both controls are fenced inside general cone sets. The authors use a change-of-formula trick and a squared-completion method to derive the best value and the best feedback rule in explicit form. The answer rests on coupled stochastic Riccati equations with jumps. They prove a unique non-negative solution exists, aided by new comparison theorems for such equations. The result gives a full recipe for a hard class of control problems.
Catalog record
| CREATOR | Xiaomin Shi, Zuo Quan Xu |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-19100
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Constrained stochastic linear quadratic control under regime switching with controlled jump size Xiaomin Shi, Zuo Quan Xu Abstract . In this paper, we examine a stochastic linear-quadratic control problem characterized by regime switching and Poisson jumps. All the coefficients in the problem are random processes adapted to the filtration generated by Brownian motion and the Poisson random measure for each given regime. The model incorporates two distinct types of controls: the first is a conventional control that appears in the continuous diffusion component, while the second is an unconventional control, dependent on the variable z z , which influences the jump size in the jump diffusion component. Both controls are constrained within general closed cones. By employing the Meyer-Itô formula in conjunction with a generalized squares completion technique, we rigorously and explicitly derive the optimal value and optimal feedback control. These depend on solutions to certain multi-dimensional fully coupled stochastic Riccati equations, which are essentially backward stochastic differential equations with jumps (BSDEJs). We establish the existence of a unique nonnegative solution to the BSDEJs. One of the major tools used in the proof is the newly established comparison theorems for multidimensional BSDEJs. Keywords: linear-quadratic control; regime switching; controlled jump size; fully coupled stochastic Riccati equations; backward stochastic differential equations with jumps. Mathematics Subject Classification (2020): 60H30. 60J28. 60J76. 93E20. 1. Introduction Since the pioneering work of Wonham [ 22 ] , stochastic linear-quadratic (LQ) theory has been extensively studied by numerous researchers. For instance, Bismut [ 2 ] was the first one who studied stochastic LQ problems with random coefficients. In order to obtain the optimal random feedback control, he formally derived a stochastic Riccati equation (SRE). But he could not solve the SRE in the general case. It is Kohlmann and Tang [ 12 ] , for the first time, that established the existence and uniqueness of the one-dimensional SRE. Tang [ 18 , 19 ] made another breakthrough and proved the existence and uniqueness of the matrix valued SRE with uniformly positive control weighting matrix using two different approaches. Sun, Xiong and Yong [ 17 ] studied the indefinite stochastic LQ problem with random coefficients. Hu and Zhou [ 10 ] solved the stochastic LQ problem with cone control constraint. Zhang, Dong and Meng [ 25 ] made a great progress in solving stochastic LQ control and related SRE with jumps with uniformly definite control weight by inverse flow technique. Li, Wu and Yu [ 13 ] considered the stochastic LQ problem with jumps in the indefinite case. Please refer to Chapter 6 in Yong and Zhou [ 24 ] for a systematic account on this subject. Stochastic LQ problems for Markovian regime switching system were studied in Wen, Li and Xiong [ 21 ] and Zhang, Li and Xiong [ 26 ] where weak closed-loop solvability, open-loop solvability and closed-loop solvability were established. But the coefficients are assumed to be deterministic functions of time t for each given regime i in the above papers, so their SREs are indeed deterministic ordinary differential equations (ODEs). Hu, Shi and Xu [ 6 , 7 ] formulated cone-constrained stochastic LQ problems with regime switching on finite time horizon and infinite time horizon respectively, in which the coefficients are stochastic processes adapted to the filtration generated by the Brownian motion for each give regime i . Due to the randomness of the coefficients, the corresponding SREs in [ 6 , 7 ] are actually BSDEs. In this paper, we generalize the LQ problem in [ 6 ] to a model in which the coefficients are stochastic processes adapted to the filtration generated by the Brownian motion and the Poisson random measure for each give regime i . In addition to a usual control u_{1} , we introduce a second control u_{2}(z) depending on the jump size z . The motivations to incorporating the second control are, in insurance area, the optimal reinsurance strategies may depend on the claim size in general, see, e.g., Liu and Ma [ 14 ] and Wu, Shen, Zhang and Ding [ 23 ] ; and in controllability issues for stochastic systems with jump diffusions, a control depending on the jump size is necessary as a consequence of martingale representation theorem of Poisson random measures, see, e.g., Goreac [ 5 ] and Song [ 16 ] . The application of this kind of stochastic LQ model in a optimal liquidation problem with dark pools can be found in our working paper [ 3 ] . The first main contribution of this paper is to provide a pure analysis method (using tools like approximation technique, comparison theorem for multi-dimensional BSDEJs, log transformation, etc) of the existence of a unique solution to the corresponding system of SREs, which is a 2\ell -dimensional coupled BSDEJs. This is interesting in its own right from the point of view of BSDE theory. Note that even though the SREs in [ 6 ] are 2\ell -dimensional, they are partially coupled, that is, the first \ell equations for \{P^{i}_{1}\}_{i\in\mathcal{M}} and the second \ell equations for \{P^{i}_{2}\}_{i\in\mathcal{M}} are totally decoupled. But in our new model, the equation for P^{i}_{1} also depends on P^{i}_{2},\Gamma^{i}_{2} , rendering the 2\ell -dimensional SREs in our new model are fully coupled. This more complicated phenomenon comes from the fact that, to the best of our knowledge, the optimal state process will probably change its sign at the jump time of the underlying Poisson random measure. Compared with the 2 -dimensional SREs in Hu, Shi and Xu [ 9 ] , here we need to study 2\ell -dimensional SREs because of the new coupling terms \sum_{j}q^{ij}P^{j}_{1} and \sum_{j}q^{ij}P^{j}_{2} . The second main contribution is to give a rigorous verification theorem of the optimal value and optimal control, using the unique solution to the corresponding system of SREs, Meyer-Itô’s formula, a generalized squares completion technique and some delicate analysis. The rest part of this paper is organized as follows. In Section 2 , we formulate a constrained stochastic LQ control problem with regime switching, controlled jump size and random coefficients. Section 3 is devoted to proving the existence of a unique nonnegative solution to the related 2\ell -dimensional fully coupled SREs in standard and singular cases. In Section 4 , we solve the LQ problem by establishing a rigorous verification theorem. 2. Problem formulation Let (\Omega,\mathcal{F},\mathbb{F},\mathbb{P}) be a fixed complete filtered probability space. The filtration \mathbb{F}=\{\mathcal{F}_{t},t\geq 0\} is generated by the following three independent random sources augmented by all the \mathbb{P} -null sets. • The first random source is a standard n -dimensional Brownian motion W_{t}=(W_{1,t},\ldots,W_{n_{1},t})^{\top} . • The second one is an n_{2} -dimensional Poisson random measure N=(N_{1},\ldots,N_{n_{2}})^{\top} defined on \mathbb{R}_{+}\times\mathcal{Z} , where \mathcal{Z}\subset\mathbb{R}^{\ell}\setminus\{0\} is a nonempty Borel subset of some Euclidean space. For each k=1,\ldots,n_{2} , N_{k} posses the same stationary compensator (intensity measure) \nu(\operatorname{d}\!z)\operatorname{d}\!t satisfying \nu(\mathcal{Z})<\infty . The compensated Poisson random measure is denoted by \tilde{N}(\operatorname{d}\!t,\operatorname{d}\!z) . • The third one is a continuous-time stationary Markov chain \alpha_{t} valued in a finite state space \mathcal{M}=\{1,2,\ldots,\ell\} with \ell\geq 1 . The Markov chain has a generator Q=(q_{ij})_{\ell\times\ell} with q_{ij}\geq 0 for i\neq j and \sum_{j=1}^{\ell}=0 for every i\in\mathcal{M} . Besides the filtration \mathbb{F} , we will often use the filtration