Whole-body momentum control with linear quadratic state incremental walking pattern generation and a centroidal moment pivot balancing strategy for humanoid robots

Abstract

In this paper, a proposed momentum control framework adopts the quadratic programming method to handle the multi-task prioritized problem by assigning suitable constraints and the objective function, including desired end-effectors motions, the limitations of joints and the rate of change of linear and angular momentum of a robot. First, the centroidal momentum matrix is used to derive the relationship between the linear and angular momentum of the center of mass and the whole-body motion of the robot. Second, the linear quadratic state incremental walking pattern generator is used to regulate the rate of change of linear momentum of the robot for balanced walking. Then, according to the difference between the centroidal moment pivot and the zero moment point (ZMP), the compensatory rate of change of horizontal angular momentum can be calculated to keep the balance from unexpected larger disturbances. Finally, the performance of this control framework is verified effectively through the simulations and the experiments. In addition, the ZMP tracking performance is improved up to 50% in the simulations and 14% in the experiments.