Stochastic programs without duality gaps (Q1925782)

The paper studies convex dynamic stochastic optimization problems in presence of an additional random perturbation parameter. The optimal value function is convex and sufficient conditions for its lower semicontinuity are derived. These conditions imply the existence of solutions and the absence of a duality gap. The proof uses extended dynamic programming equations. An application to an optimal consumption problem and a generalization of certain no-arbitrage conditions from mathematical finance are discussed.