Various strategies and algorithms have been developed for control of autonomous agents in a fleet. Many of these approaches are interesting but have limited performances in environments where measurements are less accurate or faults occur. The main contribution of the paper consists in the design of a robust flocking algorithm with regards to the problem of defective agents. Our approach is based on an extension of proposed models in literature, in order to synthesis a robust and suitable model for outdoor fleet control. The method is tested in simulation on a fleet of mobile robots. Test results and comparisons between different simulation requirements are presented.