The objective of this paper is the definition of a distributed model predictive control scheme for freeway systems. The proposed control scheme is based on an innovative reformulation of the adopted prediction model. The control approach is stated in a centralized framework, in two different cluster-based distributed architectures and in a decentralized one. Specifically, control distribution is exploited in this paper as a means to increase the practical applicability of an advanced control scheme that is often limited, due to its computational complexity and to the typical size of the systems under concern. The novelty of the proposed approach mainly relies on the way in which the control distribution is operated, since it is dictated by the subdivision of the freeway system into portions, possibly with heterogeneous features and of different sizes, called clusters. A cluster, in our view, is a system , so that the whole freeway can be regarded as a system of systems. The performance of the control scheme in its different versions is assessed in simulation, making reference to a case study developed on the basis of real data.