This paper exploits the concept of periodic invariance in order to derive a novel predictive control algorithm for switched systems. The offline computation here concerns the definition of sequences of switches which return the state vector to a set after a given number of moves, while an online optimization is used to improve performance by exploiting information available on the value of the state at each time instant. The algorithm is shown to have closed loop stability and to ensure that the state is steered to a bounded set of ellipsoids centered on the target state. The results of the paper are illustrated by means of a numerical example. (C) 2009 Elsevier Ltd. All rights reserved.