This paper presents an experimental study of robustness of multi-objective optimal sliding mode control, which is designed in a previous study. Inertial and stiffness uncertainties are introduced to a two degrees-of-freedom (DOF) under-actuated rotary flexible joint system. A randomly selected design from the Pareto set of multi-objective optimal sliding mode controls is used in the experiments. Three indices are introduced to evaluate the performance variation of the tracking control in the presence of uncertainties. We have found that the multi-objective optimal sliding mode control is quite robust against the inertial and stiffness uncertainties in terms of maintaining the stability and delivering satisfactory tracking performance as compared to the control of the nominal system, even when the uncertainty is not a small quantity. Furthermore, we have studied the effect of upper bounds of the model estimation error on the stability of the closed-loop system.