In this paper, the continuous-time optimal stationary control theory is employed to model the human motor control system in the presence of signal-dependent noise. The control scheme is based on Phillis's earlier work , and is simulated and compared with human data. The control model is compatible with some important observations, such as the asymmetry of the velocity profiles and the overshoot phenomenon. The infinite-horizon optimal control model advocated here differs from previous methods in not requiring a priori knowledge of a finite stopping time, but focuses on minimization of the variance in steady state. It is hypothesized that selecting appropriate weighting matrices in the cost function plays a key role in planning arm movement trajectories in the central nervous system, while the stopping time is not necessary to be known in planning movements.