A model predictive control strategy for finite-time reference synchronization in multi-agent systems with double-integrator dynamics