TABLE OF CONTENTS The concept of a Central Pattern Generator (CPG), a network of neurons producing the locomotion pattern within a lattice of neural activity, is a suitable paradigm for biologically inspired control of artificial locomotion. In this Chapter a new approach to the design of CPGs, based on Cellular Neural Networks (CNNs), is presented. Prom a biological point of view this new approach includes an approximated chemical synapse implemented in a CNN structure. This allows us to obtain a general class of artificial CPGs in which the desired locomotion pattern and switching among patterns are achieved by means of a spatio-temporal algorithm implemented in the CNN structure.
Walking animals employ several distinct periodic patterns of leg movements, called gaits. Stereotyped movements (locomotion patterns) are also observed in swimming, flying, crawling animals. The hypothesis underlying the CPG is that these patterns are due to a pattern of neural activities within the CPG network (the key feature of these networks is that the pattern of neural activity is mapped onto the pattern of locomotor activity). This concept is useful in robotics, since the control of a legged robot is a hard task, involving a high number of interconnected degrees of freedom (DOF). Its application to robotics gives rise to the following problem: given a desired behavior (i.e. a locomotion pattern), generating the feedforward signals driving the legs of the robot.
A widespread approach to modelling CPGs is to use dynamical systems [Cohen et. al. (1982); Collins and Stewart (1993a); Collins and...
