Stochastic control for a class of nonlinear kernels and applications (Q1747758)

The paper deals with a stochastic control for a class of nonlinear kernels. It consists of two parts, a theoretical one and applications. The first part, composed of Sections 2 and 3, contains a control on a class of nonlinear stochastic kernel and the dynamic programming principle and next path regularization of the value function. This part of the paper has a strictly theoretical character. The main result of Section 2 and the article, formulated as Theorem 2.1, is the dynamic programming principle (DPP in short) proved in an abstract setting. The DPP remains an essential tool in the control theory. This principle states that a global optimization problem can be split into a series of local optimization problems. Unfortunately, the proof of DPP is generally tough and complicated. There are several approaches to DPP. In their proof, the authors generalize the measurable selection argument to derive DPP in the context of optimal stochastic control of nonlinear expectations (kernels), which can be represented by backward stochastic differential equations (BSDEs for short). Next in Section 3, a semimartingale decomposition for the value function of the control problem is given. The second, remaining part of the paper supplies us several applications of the obtained theoretical results. In Section 4, the authors prove a well-posedness result for second order BSDEs, the uniqueness of the solution, comparison results, estimates for 2BSDEs and some estimates for the difference of two solutions to 2BSDEs. See Theorems 4.1, 4.2, 4.3, 4.4 and 4.5, respectively. In Section 5, a nonlinear and robust generalization of the so-called optional decomposition for supermartingales is obtained. Theorem 5.1 provides the existence and uniqueness of the solution to saturated 2BSDEs. Theorem 5.2 extends results for continuous processes to markets with nonlinear portfolio dynamics. It gives a super-hedging strategy for the terminal condition. Finally in Section 6, the authors recall the link between 2BSDEs and path-dependent PDEs under some additional regularity assumptions. Theorems 6.1 and 6.2 give the characterization of the value function as a viscosity solution of PPDE. The paper finishes with the appendix containing technical results for BSDEs and 87 references. It is worth to emphasize that results obtained in the article are essential not only for people working in BSDEs but for people interested in finance, too. This large paper is very well written in details. Additionally, its interesting introduction presents a history and several aspects of the problems considered.