Geometry of distribution-constrained optimal stopping problems (Q1626603)

This paper is about an optimisation problem regarding stopping times over a given filtered probability space that have a given distribution. Given an adapted process \(B\) and a cost function the aim is to find the stopping time \(\tau\) with distribution \(\mu\) that minimises the expected value of \(c(B,\tau)\). The authors show that if this cost function is lower semicontinuous and bounded from below, then the problem has a solution, in the case where \(B\) is a Brownian motion starting a point which has a given distribution and \(\mu\) has finite \(p_0\) norm for some \(p_0\geq 0\). Their second result is a monotonicity principle. According to this, there exists a set \(\Gamma\) of paths and times (associated with each path) such that the Brownian paths at the values of the optimal stopping time belong to \(\Gamma\) with probability 1 and furthermore for all pairs of paths and times \((\omega,t) \in \Gamma\) the expected cost of stopping earlier than time \(t\) is higher for any stopping time \(\sigma\) of finite \(p_0\) norm.