One of the challenging problems with autonomous vehicles is their performance at intersections. This paper shows an alternative control method for the coordination of autonomous vehicles at intersections. The proposed approach is grounded in multi-robot coordination and it also takes into account vehicle dynamics as well as realistic communication constraints. The existing concept of decentralized navigation functions is combined with a sensing model and a crossing strategy is developed. It is shown that, thanks to the proposed approach, vehicles have smoother trajectories when crossing at a four-way intersection. The proposed method is compared to adaptive traffic lights and roundabouts in terms of throughput. Results show that using a decentralized navigation function for the coordination of autonomous vehicles improves the performance by reducing energy consumption and pollution emission.