A lesion study of a heterogeneous artificial neural network for hexapod locomotion

Abstract

An artificial neural network for hexapod locomotion that is extremely robust, in large part because of the way in which it functions, has been constructed. Both its robustness and its function are due to the dynamic interactions between its central and peripheral components. However, the relative roles of these two components differ at different speeds. Faster gaits are primarily generated centrally, whereas slower gaits are more dependent on sensor information. In all gaits disruptions of sensory information can often be compensated for by central information, and vice versa. These results demonstrate the utility of heterogeneous neural networks for the generation of complex behavior. They also shed light on the relative roles of central and peripheral mechanisms in the generation of patterns of behavior in biological nervous systems. They show that lesion studies provide a powerful method for probing the dynamics of heterogeneous neural networks.