Spinal Cord Like Artificial Neural Networks and Application to Robots

Abstract

In general, robot movements are realized by computational processing that combines microprocessors and software. However, actual living things do not have such CPUs and software, and various processing is performed by combining neural networks. In processing related to movement, nerve signals are generated mainly in the spinal cord and transmitted to the motor nervous system. Organisms perform diverse and flexible processing by combining relatively simple neurons. So far, several studies have been reported on generating robot motions by forming artificial neural networks, but most of them are based on software calculations. We take the position of using hardware to shape the spinal cord. In this paper, we describe the configuration of the hardware spinal cord, apply it to a microrobot and a bipedal robot, and show that we have realized gait modification that changes the pattern of each gait. In this study, a signal pulse was input from the upper central nervous system to an artificial spinal cord consisting of a hardware neural network, and the pulse pattern of the CPG was modified accordingly. Specifically, we controlled gait changes in quadrupedal and bipedal locomotion. This means that among the functions of the voluntary and involuntary systems involving the superior central nervous system, the rhythm that promotes walking was generated and changed.