Simulator of mobile robots controlled by artificial neural networks to learning courses in robotics

Abstract

A mobile robot simulator was developed in a computer graphics environment as a didactic tool able to assist the teaching and learning in disciplines that involve mobile robotics. It is also useful in validating control algorithms before testing on a real robot. The developed interface allows the user to adjust the parameters of the robot, the controller and exchange maps with different obstacles. In addition to the kinematics and 2D modeling of the virtual mobile robot with wheels, three sensors were used that indicates the robot's distance to obstacles and walls. The robot controller uses a Multilayer Perceptron Neural Network (MLPNN) for autonomous navigation avoiding the collision with obstacles.