Using a geometric Brownian motion to control a Brownian motion and vice versa (Q1275935): Difference between revisions

Consider a one-dimensional controlled process \(x(t)\) governed by the equation \[ dx(t)=a(\xi (t))dt+b(\xi (t))u(\xi (t))dt+[N(\xi (t))]^{1/2}dW(t), \] where \(\xi (t):=(x(t),t)\). The aim of the homing control problem is to minimize the expectation of a functional of the form \[ J(x)=\int _0^{T(x)} [\tfrac {1}{2}q(\xi (t))u^2(\xi (t))+\lambda ]dt, \] where \(q\geq 0,\;\lambda \) is real and \(T(x)\) denotes the exit time from an interval \((A,B)\) for a solution starting from \(x=x(0)\in (A,B)\). In the particular case \(a=0,\;b=N=1\) and \(q(\xi (t))=x^2(t)\) the optimal control is found by means of the mathematical expectation of a geometric Brownian motion while the optimal process is shown to be a Bessel process. Conversely, if the uncontrolled process is a geometric Brownian motion, then the optimal control is found by means of an expectation of a Brownian motion.