Improving ZMP-based control model using system identification techniques

The approximation of humanoid robot by an inverted pendulum is one of the most used model to generate a stable motion using a planned Zero Moment Point (ZMP) trajectory. In this paper, we aim at proposing to improve the reliability of this model using system identification techniques. To achieve this goal, we propose an identification method which is the result of the comprehensive application of system identification to dynamic systems. Moreover, we propose a controlling algorithm for the identified model in oder to track a desired trajectory of ZMP. The efficiency of the method is shown using dynamical simulation and conducting real experiments on the humanoid robot HRP-4C.