When employing autonomous wheeled robots, it is desirable to use navigation approaches which can always prevent collisions. In this paper we consider the problem of navigating multiple vehicles through an unknown static environment with limited sensing and communication capability available. We propose a decentralized, cooperative, reactive, model predictive control based collision avoidance scheme, and show it is able to prevent collisions from occurring. An axillary controller is employed to follow previously planned paths when the main path planning system fails to find a path. Simulations in various scenarios confirm the methods validity.