We propose a cooperative adaptive cruise control which exploits a front vehicle acceleration forecast, available via vehicle-to-vehicle communication, to reduce vehicle energy consumption. The proposed control design can be applied to groups or platoons of cooperative vehicles, which locally optimize performance while maintaining their relative distance bounded at all times. The proposed approach adopts a learning-based model predictive control scheme, and maximizes performance while guaranteeing robust safety against the uncertainty on the front vehicle acceleration forecast. Simulations show the validity of the proposed control design.