Playing a tic-tac-toe game against a robot manipulator using a vision system

Abstract

This work deals with the integration of Robotics, Artificial Intelligence and Image Processing. To do that, a experimental setup was assembled at the Robotics Laboratory of the Mechanical Engineering Department of the Taubate's University, composed by a CCD black and white camera, a frame grabber with 8 bits resolution and a didactic robot kit Robix RCS-6. The main idea is to implement a tic-tac-toe game where the robot manipulator plays against a human opponent. The camera scene is composed by the tic-tac-toe board and the robot can ";see"; it using Image Processing techniques. The board's image is acquired repeatedly and an image subtraction between two simultaneous images is done. The threshold of the resultant image is calculated and a binary image is obtained. The centroid of the remain figure is calculated and used to allows the robot decides in which board cell it should play to block its opponent. Once this decision is taken, the robot, assembled in a two degree of freedom configuration, draws an X on the game board. Algorithms for threshold calculus and the robot inverse kinematics are discussed. The scene illumination and problems tha it address are also discussed. The program, developed in C++ language, works quite well and illustrate how those areas (Robotics, Image Processing and Artificial Intelligence) can work together.