The issue investigated in this paper concerns navigation, survey / control and the complexity associated with a mobile multi-robot coordination and cooperation in a complex environment (robotic analysis laboratory), which is little or no known with significant industrial risks, in the presence of human and machines. This group of mobile robots is mainly used to move chemical products, which can lead dangerous accidents (toxic, flammable, explosive …) between the different rooms of the laboratory. The objective of our study is to ensure a good precision in the robots navigation in order to optimize human efforts, reduced error and establishment safety while keeping an eye on robots with good functioning and a desired production. In the literature there are several risk analysis techniques. Among the most used techniques in robotics, the FMEA method (failure modes, effects and criticality analysis). We applied the FMEA method on one robot. Then, the FTA (Fault Tree Analysis) method was chosen to generalize dependability study on all robots. Finally, to manage this level of complexity, a control architecture based on controllers' decomposition into a set of elementary behaviors / controllers (obstacles avoidance and collision between robots, attraction to a target, planning …) was proposed.