Biped Walking Control Based on Quadratic Programming and Differential Inequalities

Abstract

This paper presents a novel method to control a bipedal walking based on quadratic programming and differential inequalities using geometric primitives. We allow the center of mass to move anywhere inside the support polygon during the walking cycle, as opposed to classic methods, which usually rely on tracking a desired trajectory for the zero moment point. The constraints keep the robot balance, the pelvis above a minimum height, and prevent the violation of joint limits during the complete walking cycle. Simulation results using the legs of the Poppy humanoid robot show that the trajectories of the closed-loop system converge to the desired center of mass position during the double support phase and the swing foot's trajectories converge to the desired pose during the single support phase while all constraints are obeyed.