An Improved Method Based on CPG to Achieve Trotting Gait of Quadrupedal Robot

Abstract

In this paper, we achieve a trotting gait of quadrupedal robot based on neural oscillator networks in simulation. Through combining the output signals of neural oscillator networks and feedback signals derived from the pose of body of quadrupedal robot, we make the quadruped travel with dynamical stability. Compared with previous researches in this way, we abandon the assumption that, when leg is in stance phase, only the upper limb of leg swings backward and the lower limb still keeps the original state (corresponding to the state that leg just finished the swing phase). Consequently, the amplitude of pitch and roll motions of torso of quadrupedal robot is greatly attenuated. By the modified method, a more steady motion with quadrupedal robot is achieved.