We investigate an optimal control of a system with a specific form of hysteresis. The hysteresis operator considered in the article is a Play operator which generalizes the usual Play operator defined through a one dimensional domain set. The main novelty of this paper lies in the study of the Play operator with a set of feasible states which is multidimensional and possibly nonconvex. The key ingredient of our approach is the interpretation of the hysteresis control systems through a reflected control system. We provide a characterization of the value in terms of the unique solution to a Hamilton-Jacobi-Bellman variational inequality.