REALIZATION OF NEURAL NETWORK FOR GAIT CHARACTERIZATION OF QUADRUPED LOCOMOTION

Abstract

Animal central pattern generator (CPG) is a device used to simulate the nervous system of animals. It is widely used in the design and control of four-legged robots. The objective of this paper is to establish a CPG network which is formed by a set of mutually symmetric neural networks combined with time delay to generate rhythmic motion patterns. Firstly, a symmetric delayed neural network consists of two loops and composed of eight neurons that can produce multi-phase locked oscillation patterns corresponding to the quadruped gaits. Then the primary gaits of all six types can be produced,and gait transitions between the different gaits are generated by altering the delay as the parameter, i.e., different ranges of delay correspond to different patterns of neural neurons. At last£¬ the simulation results show that the delayed neural network can generate multiple periodic oscillations corresponding to the gait of quadruped locomotion.