Gait generation and transitions of quadruped robot based on Wilson-Cowan weakly neural networks

Abstract

This paper presents a new Central Pattern Generator (CPG) controller for quadruped robot rhythmic locomotion control. Our CPG controller based on the Wilson-Cowan neural oscillator, which is known as a weakly neural network that generates fundamental rhythmic movements in locomotion of animals. The harmony motion of one leg from the others is controlled with four Wilson-Cowan neural oscillators. By replacing the gait matrix of neural oscillators, some different rhythmic locomotion models and switching between the models can be generated. The parameter normalization is introduced for controlling the output amplitude of the CPG easily. Simulation results show that the presented CPG controller can generate different quadruped gaits and change its rhythmic patterns smoothly. The period and amplitude of the CPG model are easy to control for generating various gaits.