This work proposes a decentralized control architecture for a group of car-like vehicles. A network of two degree decentralized controllers is used to stabilize the robots in its desired position. The controllers first generate an on-line feasible optimal or suboptimal trajectory, then a feedback compensation is performed around this trajectory. To generate the optimal trajectory, two reactive terms are minimized adapting this one to avoid the obstacles and assure the communication links between the robots. The proposed architecture is supported by a global positioning system GPS, and wireless communication for each robot.