In order to handle the model uncertainty and the external disturbance, a robust attitude control method for helicopter robots is proposed in this paper. Unscented Kalman Filter (UKF) and backstepping technique are adopted in the attitude control design. Model-based backstepping control is presented to keep the desired helicopter attitude. UKF is employed for online estimation of both motion states and model errors of the helicopter. Such estimation results are further incorporated into the controller of helicopter. The backstepping control enhanced by real time model errors feedback can achieve a robust tracking performance even with the occurrence of external disturbance. Simulations conducted on the helicopter illustrate the effectiveness of the proposed attitude control.