Reducing a nonlinear dynamic programming equation to a kolomogorov equation (Q4372872)

The one-dimensional dynamic system defined by the stochastic differential equation \[ dx(t) =f(x(t))dt + B(x(t))u(t)dt + [N(x(t)]^{1/2}dW(t), \] where \(x(t)\) is the state variable, \(f(\cdot)\) is an arbitrary function, \(B(\cdot)\) is different from zero, \(u(t)\) is the control variable, \(N(\cdot)\) is positive and \(W(t)\) is a standard Brownian motion, is considered. A theorem that enables to compute the optimal control of a diffusion process from a mathematical expectation for the uncontrolled process when a certain relation holds is generalized. The optimal control is obtained from a mathematical expectation for the uncontrolled (or another) process, even if the relation in question is not satisfied. The cost criterion takes the risk-sensitivity of the optimizer into account. The example of the stochastic differential equation \(dx(t)=f(x(t))dt+x(t)u(t)dt+\sqrt{x(t)}dW(t)\) is presented.