A constrained non-linear regular-singular stochastic control problem, with applications. (Q2574623)

The controlled process \(X(\cdot ) :=\{X(t): 0\leq t<\infty \}\) governed by the stochastic differential equation \[ dX(t) =[\mu U(t)-aU^2(t)-\delta ]\,dt + \sigma U(t)\,dW_t -dZ_t,\quad X(0-)=x, \] is considered, where \(\mu , a, \sigma ,\delta \) are given constants, \(W_t\) is a standard scalar Wiener process, \(Z_t\) is a non-negative, non-decreasing, adapted process that is right-continuous with left limits, and \(U\) is an adapted process constrained by \(l\leq U(t)\leq u\) for some constants \(l\) and \(u\). The aim is to find an optimal admissible control \((U^*(t),Z^*_t)\) that maximizes the cost \[ E\int _0^\tau e^{-rt}\,dZ_t, \] where \(\tau \) is the first hitting time of \(\{0\}\) over all admissible controls. The value function and optimal controls are found in a closed form by solving the corresponding HJB equation. Some applications are also discussed.