Robot training using neural networks

Cooperative work of self-organizing maps (SOM) and multi layer perceptron (MLP), which are kinds of neural network models, on a mobile robot was examined. A microprocessor controlled mobile robot that has high movement capacity, ultrasonic and light intensity sensors on it has wireless communication feature using Bluetooth was built. Mobile robot tracking in a maze to a target signed with a light source was planned in a less time and with keeping away from walls using neural network models. SOM is used to keep the robot away from walls and MLP is used to determine the movement of mobile robot to travel to the target. Neural network models were built on a personal computer using MATLAB Neural network toolbox. In this work the results of mobile robot tracking to the target before and after neural network training was examined and compared. Keywords: Robot, self-organizing maps (SOM), multi layer perceptron (MLP), Bluetooth, ultrasonic sensors, neural network.