Martingale selection problem in the case of finite disrete time

We consider a multivalued stochastic process specified on a filtered probability space. Assuming that the values of the process are convex we establish a criterion for the existence of an adapted sequence of selectors that can be transformed into a martingale by an equivalent change of measure. The criterion has a geometric nature and is expressed in terms of the supports of the regular conditional upper distributions of the elements of the multivalued process.