Hardware neural network models of CPG and PWM for controlling servomotor system in quadruped robot

This paper discusses the pulse-type hardware neural networks (P-HNNs) that contain a central pattern generator (CPG) and a pulsewidth modulation (PWM) servomotor controller and the application to quadruped robots. The purpose of our study is mimicking the biological neural networks and reproducing the similar motion of the living organisms in the robot. The CPG of the living organism generates the walking rhythms. We mimicked this CPG by modeling the cell body and the synapse of the living organism. The developed CPG composed of the P-HNN output four pulse signal sequences and the four outputs are introduced to each leg of the quadruped robot. On the other hand, the angle of the servomotor is controlled by the PWM. The PWM is obtained by modeling the axon of the living organism. The CPG and the PWM servo control system perform the walking motion of the quadruped robot. Moreover, the gate pattern change of quadruped animals is reproduced by these P-HNNs.