Abstract
In this paper we introduce a notion of causal transference plans on Polish spaces. These plans are probabilities on the product space which generalize the adapted processes of stochastic calculus in the same way as the Monge-Kantorovich transference plans generalize the Monge transference plans. We provide a detailed study of their main properties. Then we introduce the associated causal transport problems and we prove a very general result of existence of solutions to the causal Monge-Kantorovich problems. Finally we relate these problems to stochastic optimal control, and we investigate the transports of the Wiener measure for the quadratic cost. Weak solutions to some stochastic differential equations whose solutions can be obtained by transformation of the drift then appear as optimal transference plans to these causal Monge-Kantorovich problems, while the existence of a unique strong solution to these equations is related to the existence of an optimum of Monge. In this case, the causal counterpart of the Wasserstein distance is the square root of the relative entropy.

This publication has 24 references indexed in Scilit: