Four-legged robot's behavior controlled by pulsed para-neural networks (PPNN)

Abstract

This paper deals with the synthesis of behaviors of a motor-driven four-legged robot controlled by Pulsed Para-Neural Networks (PPNN). During the first step of the synthesis, based on a simulation model of the robot and a description of its desired behavior, the required function of torque versus time was generated for each motor of interest. Then, for each Frequency-to-Voltage Converter (FTVC) associated with given motor, a required spiketrain is generated. After this, a PPNN is then created in a 3-D Cellular Automata space. We present principles and a software tool ("NeuroMaze") for PPNN synthesis. As a sample implementation, we discuss a robo-kitten and the PPNN responsible for the execution of the actions required to catch a robo-mouse.