Our model is used for a simulation in the framework of mobile robotics and more specifically collective robotics. Our application tackles a quite common problem in collective robotics which is still given a lot of consideration: agents are in charge of regrouping objects distributed in their environment. The innovative aspect of our approach rests indeed on a system integrating operationally autonomous agents: every agent in the system has the freedom to act on a cell (the agent decides by itself which action to take). Therefore, there is not in the system any type of master responsible for supervising the agents, nor any type of cooperation protocol, thus allowing the system to be more flexible and fault tolerant.
We implemented a serial version of this simulation in the Swarm Simulation System [Langton et al. 1996], stressing on the measurement of quantitative results and on the realization of appropriate visualization tools to follow in real-time the run of an experiment. We implemented several variants for agent modules and realized an intensive number of experiments whereof we observed an implicit cooperation between the agents in the system to accomplish a global task, i.e. regrouping objects. In such an experiment, the location of the stack containing the objects at the end of the run is the result of the agents' interactions. Details on the implementation and results can be found in [F. Chantemargue et al. 1996] and [Dagaeff et al. 1997]. A preliminary implementation of this simulation in a real world involving real mobile robots (kheperas) exhibited the emergence of cooperation.