Biologically Inspired Neural Controller for Robot Learning and Mapping

Abstract

In this paper we present a model composed of layers of neurons designed on the basis of the neurophysiology of the rat hippocampus to control the navigation of a real robot. The model allows the robot to learn reward locations in different mazes and to return home autonomously by building a topological map of the environment. We described robotic experimentation results from our tests in a T-maze, an 8-arm radial maze and a 3-T shaped maze.