The advent of connected, automated and autonomous vehicles introduces the possibility for new traffic control approaches. Vehicles equipped with automation and communication systems can be exploited both as sensor and actuators for traffic control actions, thus avoiding the need for new infrastructure. In this paper a multi-class extension of the macroscopic Cell Transmission Model is adopted to describe the interaction between different classes of vehicles, for example human-driven and connected/automated. The vehicle classes are distinguished on the basis of their time headways and their speed. By means of a Model Predictive Control approach, the optimal free-flow speed for the class of connected/automated vehicles is computed and applied to them with the aim of reducing congestion on the highway. The effectiveness of the proposed control law is analyzed depending on the penetration rate of controlled vehicles and the approach is assessed in simulations.