MATHEMATICS

On deterministic approach to solution of stochastic optimal control problem with controlled diffusion

N. S. Ismagilov

Department of Mathematics, Ufa State Aviation Technical University, ul. K. Marksa, 12, Ufa, 450000, Russia

Abstract: We consider an optimal control problem for a one-dimensional process driven by stochastic differential equation, which has both drift and diffusion coefficients controlled, diffusion being linear in control
\begin{equation} dx(t) = b(t,x(t),u(t))\,dt + \sigma(t,x(t))u(t)\,dW(t), \quad x(0) = x_0, \nonumber \end{equation}
where $x(t)$ is the state variable, $u(t)$ is the control variable and $W(t)$ is the Wiener process. We prove a theorem which gives a structure of solution for the considered differential equation as a superposition of functions $x(t) = \Phi(t,u(t)W(t) + y(t))$, where $\Phi(t,v)$ is the known function, which is completely determined by the diffusion coefficient $\sigma(t,x)$ and is independent of control, and $y(t)$ is the solution to the pathwise-deterministic measure-driven differential equation
\begin{equation} dy(t) = B(t,y(t),u(t))\,dt - W(t)\,du(t). \nonumber \end{equation}
The revealed structure of the solution enables us to consider a new pathwise-deterministic impulsive optimal control problem with the state variable $y(t)$ which is equivalent to the original stochastic optimal control problem. Pathwise problems may have anticipative solutions, so we propose a method that makes it possible to build nonanticipative optimal solutions. The basic idea of the method is to modify cost functional in new pathwise problem with special integral term, which guarantees nonanticipativity of solutions.

Keywords: stochastic optimal control, stochastic differential equations, deterministic approach, pathwise optimization, optimal impulsive control.

UDC: 519.21, 517.977

MSC: 93E20, 49K45, 60H30, 49N25

Received: 29.10.2013

DOI: 10.20537/vm140202