In this work, the Acromovi architecture has been explained. This architecture is the support for all the programming and interaction of the agents that manages our team of robots. The Acromovi architecture is basically an agent-based distributed architecture for the programming and controlling of teams of mobile robots. This architecture allows the use of native components, by means of agent wrappers. The agents add the distributed capabilities, with message interchange for cooperation between robots. This makes possible an important reutilization of code, this being a basic principle in software engineering. Also, it allows the scalability of the system. That is, if one tested application is of interest for the whole system, it can easily converted in a new agent of the architecture, to serve as basis for new applications more complex. Moreover, it implements another concept, the resources sharing. This means that all the elements of one of the robots of the team can be easily accessed by all the other robots of the teams by means of the agents that control each of the other robots. Finally, Acromovi architecture allows the quick development of multirobot applications with the advantages of the multiagent systems. The users need very little time to develop new applications for the team of robots. In just a few weeks, a student with very low Java skills is able to program agents and develop the necessary testing procedures. In this aspect also is important the first concept of the framework, the reusability of the agents previously implemented. For these reasons, this architecture is very appropriate for the implementation and execution of tasks that require collaboration or coordination by part of the robots of a team. As future work, it has been now implemented a new application for the team of robots. This application tries to form multirobot formations. Taking the example of the robot following, robots in line formation, it is possible that the each follower robot could internally add a fixed displacement to its leader position. As can be in Fig. 11, each robot computes a virtual point instead of the leader position. These virtual points can be calculated simply displacing the leader position in the correct form. Now, the followers do not follow the leader, but these virtual points.