Feedback optimal controllers for the Heston model (Q2187328)

The study refers to the controlled stochastic system \[ dX_1=\mu X_1dt+X_1\sqrt{u{X_2}}dW_1, \] \[ dX_2 =k(\theta-X_2)dt+\sigma\sqrt{X_2}dW_2, \] for \(t\in(0,T)\) and \(X_1(0)=X^0_1\), \(X_2(0)=X^0_2\) where \(W_1\), \(W_2\) are the Brownian motions, \(X_i ^0\geq 0\) and \(\mu\), \(\kappa\), \(\sigma\), \(\theta\) are positive constants. The system extends the well-known Heston model, \(u(t)\) denoting the control function. To the system one associates a performance functional of the form \[ J(u;X_1^0,X_2^0)=E\int _0 ^TX_1^2(t)f(X_1(t),u(t))dt +Eg(X_1(T)) \] where the control \(u:[0,T]->[a,b]\), \(a,b >0\). The optimal control problem one states is: if the state of the system is given from the solution of the Heston model, find the stochastic process \(u\) that minimizes the objective functional \(J\). The optimal control is obtained by means of a dynamic programming technique. The existence of at least one martingale solution \(X=(X_1,X_2)\) for the Heston model is proved first. To the optimal control problem one associates the Hamilton-Jacobi equation, which is further reduced to a nonlinear parabolic equation with some boundary conditions. The existence and uniqueness of a weak solution to the new problem is proved. Finally, the existence of an optimal feedback controller is shown.