Predictive preview trajectory tracking control for quadruped robots with Zero Moment Point

Abstract

In this paper, we propose a control algorithm for quadruped walking robots to track the Zero Moment Point (ZMP) trajectory using Preview Control. In critical moment of trot pace, we can assume the robot as a linear inverted pendulum so that the linear inverted pendulum model can be applied to describe the dynamic behavior of our quadruped robot. A preview controller therefore is designed for this non-minimum phase system to get the optimal input of the system. ZMP reference trajectory can be decomposed into two parts. One is on the direction of movement, and the other is on the vertical direction of it. With the preview controller we can track the reference in both two directions and the optimal trajectory tracking can be achieved.