Special-Purpose Computing for Neurobiologically Inspired Networks

Recent progress in understanding the mechanisms underlying animal behavior has come from the construction biomimetic robots with neural circuit controllers and the analysis of the controller dynamics during interaction with the environment. Such robots have enabled researchers to quickly experimentally test hypotheses and develop new thoeries. What is presently needed is the capability to construct more complex control networks and understand their operation in a vehicle. We propose to develop a control system for an undulatory robot based on the simple nervous system of the lamprey. We have developed physiological models of various types of neurons and synapses that are implemented in the form of a difference equations. These models can be combined to produce very realistic patterns of spiking and bursting activity of individual neurons. The models are simple enough both to reveal the underlying dynamical mechanisms and to build a network controller within a microprocessor. We will apply the models to replicate the behavior of a simple electronic nervous system that can operate in real time. Using numerical simulations we will establish the mechanisms and patterns which enable diving, climbing, turning, avoiding obstacles, rheotaxis, and swimming in a fixed direction.