Recon.gurable Pattern Generators Using Nonlinear Electronic Circuits

Abstract

In these experiments we explore the practical implementation of pattern generating electronic circuits. The project takes advantage of fundamental mathematical arguments, based on symmetry, in order to define the circuits and coupling topologies that are used. By constructing networks of low-order dynamical neuron models, and by considering symmetries in the way the neurons are coupled, including time-shift invariant symmetries, the patterns generated by the system are reduced to a predictable few. Using the resulting patterns, the relative phases between the synchronized neurons within the network are used to define gait patterns that are similar to those found in living quadrupeds. Electronic circuits based on these observations are engineered to generate the needed analog signals for driving locomotion in an N-legged robot.