Optimal Preview Control of the Nao Biped Robot using a UKF-based State Observer

In this paper, an optimal preview controller for the Nao biped robot is developed based on a three-dimensional linear inverted pendulum model and zero moment point stability criterion. In order to estimate the unmeasurable state vector, an unscented Kalman filter state observer is proposed which guarantees robustness of the control loop through loop transfer recovery and provides asymptotic stability of the error covariance. This approach necessitates developing and utilizing forward and inverse kinematic analysis. Inverse kinematic relations are employed within the controller as an estimator of joint angles given positions of the zero moment point and center of mass trajectories. The results show effectiveness of the proposed preview controller and the observer scheme to achieve a stable walk.