This paper presents the inverse optimal design method of a nonlinear distributed consensus protocol for formation control of multiple mobile robots. Both dynamics and kinematics are considered in the protocol design. First, we propose a state transformation method to obtain a proper consensus model of a mobile robot. Then, the inverse optimal protocol is designed with respect to a meaningful cost function under the assumption of perfect angular velocity tracking. The assumption will be relaxed by extending the inverse optimal protocol using the backstepping and Lyapunov's direct methods. The numerical simulation is carried out to verify the effectiveness of the proposed method.