Nonconcave robust optimization with discrete strategies under Knightian uncertainty (Q2009179)

In the article the stochastic optimization problem \[ \sup_{H \in \{H\}} \inf_{P \in \{P\}} E^P[\Psi(H)] \] is considered. Here \(\{H\}\) is the set of control processes, e.g., \(H\) describes portfolio evolution of a trader, \(\{P\}\) is a set of probability measures modelling uncertainty and \(E^P\) denotes mathematical expectation with respect to measure \(P\). \par The optimization problems under consideration are not concave. Corresponding examples of market models are provided. They include optimal stopping and semi-static trading under Knightian uncertainty. The article establishes conditions of the existence of a maximizer \(\hat{H}\) for a class of maps \(\{\Psi\}\). In particular, these conditions include discreteness and measurability assumptions of \(\Psi\).