A fundamental and open issue pertaining to human postural sway is how to deal with the uncertain, nonlinear and time-varying nature of human motor dynamics. To address the inherent limitations of the current methods, such as PID and model-based designs, a novel active disturbance rejection concept is introduced. In this new framework, the uncertainties, nonlinearities and changes in the dynamics of the plant are treated as disturbance to be rejected. A unique disturbance rejection observer is employed to estimate it and compensate for it in real time. It is shown that the resulting new controller yields excellent performance even with significant uncertainties in the plant dynamics. Furthermore, such design strategy requires very little prior knowledge of the plant.