A Computer Vision System based on Multi-Layer Perceptrons for Controlling Mobile Robots

Computer Vision System (CVS) is an important tool for robot navigation. In this paper, a CVS is proposed for controlling mobile robots which is based on Multi-Layer Neural Networks. This system is constituted by two main modules: vision module and navigation control module. The first module processes images that can contain several objects with different colors, returning the position of one of these objects, which corresponds to a given color. The control module utilizes the results obtained from the vision module providing the robot the action to be performed. The aim of the proposed system is to enable a robot to navigate in an environment following a color object and avoiding obstacles. Some experiments in a real environment with the robot Pioneer I are presented to show the good performance of proposed system.