Forward-dynamic simulations of biomechanical movement often attempt to model exact kinematic trajectories despite the measurements demonstrating inherent motion variability. Feedback can be employed to compensate for the variability but represents unrealistically excessive neuromuscular control. A five-link, 2-D, forward-dynamic simulation of human walking with minimal feedback was implemented where-in control was achieved not in terms of joint moment but rather in equivalent terms of adjustable joint rotational stiffness. Random kinematic perturbations, applied to challenge dynamic stability, increased variability but did not disrupt successful gait. Results confirm that control of conservative elastic elements improves the realism and stability of human motion simulation.