Neural Networks for Environmental Recognition and Navigation of a Mobile Robot

Mobile robots could play a significant role in places where it is impossible for the human to work. In such environments, neural networks, instead of traditional methods, are suitable solutions to locally navigate and recognize the environment\´s subspaces. In order to learn and perform two important functions "environmental recognition " and "local navigation ", multi-layered neural networks are trained to process distance measurements received from a laser range-finder. These neural networks are a major component of the control system of a mobile robot dedicated for industrial applications. This paper will focus on a computer based design and test of a neural system, that includes three neural controllers for local navigation, and two neural networks for environmental recognition, fed off-line by a simulated model of a laser range-finder.