Extending two-dimensional human-inspired bipedal robotic walking to three dimensions through geometric reduction

Abstract

Three-dimensional humanlike bipedal walking is obtained through a hybrid control strategy which combines geometric reduction with human-inspired control. Functional Routhian reduction decouples the sagittal and coronal dynamics of a biped, thereby reducing the control design problem to sagittal motion. Experimental human kinematics data have shown that certain outputs on a human's kinematics follow a canonical human function : Human-inspired controllers are designed based on this function. The parameters of these functions are found through optimization by trying to make them as close to the human data as possible while simultaneously forming a partial hybrid zero dynamics under feedback linearization. PD control is used in these controllers to track the human functions resulting in stable walking in both two- and three-dimensional simulations.