A three-neuron controller. I

Abstract

This is the first in a sequence of papers which discuss the biological plausibility of a neural control circuit consisting of two input neurons and one output neuron. An implementation of this circuit has been shown to control a simple first order linear control system on a par with a fuzzy controller designed for the same process (Alexander 94a). The neural circuit successfully maintained the height of water in a tank under conditions of varying input flow. In this paper, we discuss the equations used in the simple neural circuit; the simple neural circuit's method of controlling first order linear systems; demonstrate the threeneuron controller's capability to solve the truck backerupper (Kosko 92) problem; and discuss the applicability of the three-neuron controller (TNC) in harder problems. We also discuss a generalized software package (FuzComp) that we are developing to assist in our studies. 1. BACKGROUND