In this study, an optimal nonlinear tracking control law is analytically developed for the semi-active suspension system by the response prediction of the quarter car model, which includes the nonlinear magneto-rheological (MR) damper model. Before this, the skyhook model optimized by the genetic algorithm is presented as a desired model to be tracked by the controller. The optimum parameters of the skyhook reference model are found by minimization of the difference between the root mean square values of acceleration obtained by the power spectral density of an actual random road excitation, and the boundary values specified by the ISO2631 standard at different effective frequencies. The effectiveness of the proposed control system is investigated in the presence of model uncertainties through simulation studies conducted by random road excitations in the time domain. The optimal property of the control law provides the possibility of limiting the input current to the MR damper, as the control input, to the physically admissible range by compromising with other conflicting objectives, that is, ride quality and handling performance. The results indicate that a satisfactory dynamic performance for the suspension system through a reduced input current to the MR damper can be achieved by the proposed control law.