We implement a. self-organized flocking behavior in a group of mobile robots and analyze its transition front all aligned state to all unaligned state. We briefly describe the robot and the simulator platform together with the observed flocking dynamics. By experimenting with robotic and numerical systems, we find that an aligned-to-unaligned phase transition can be observed in both physical and simulated robots as the noise level is increased, and that this transition depends oil the characteristics of the heading sensors. We extend the Vectorial Network Model to approximate the robot dynamics and show that it displays all equivalent phase transition. By computing analytically the critical noise value and numerically the steady state solutions of this model, we Show that the model matches well the results obtained using detailed physics-based simulations.