UC Berkeley

This thesis presents novel methods for computing optimal pre-commitment strategies in time-inconsistent optimal stochastic control and optimal stopping problems. We demonstrate how a time-inconsistent problem can often be re-written in terms of a sequential optimization problem involving the value function of a time-consistent optimal control problem in a higher-dimensional state-space. In particular, we obtain optimal pre-commitment strategies in a non-linear optimal stopping problem, in an optimal stochastic control problem involving conditional value-at-risk, and in an optimal stopping problem with a distribution constraint on the admissible stopping times. In each case, we relate the original problem to auxiliary time-consistent problems, the value functions of which may be characterized in terms of viscosity solutions of a Hamilton-Jacobi-Bellman equation.