Using Neural Networks to Navigate Robots Among Obstacles

In the paper a new method of positioning the robot in space is proposed, characterized by the application of an artificial neural network to ultrasonic sensors. The method is based on the use of a smooth neural network. The network structure is proposed and a set of input parameters is defined, which are obtained from the developed experimental installation. The neural network has been trained with the aid of an algorithm to reverse spread the error on the collected set of learning examples. Comparison of the proposed method with existing technologies that are used to solve the navigation problem. The new method allowed for the construction of maps of the depth of the surrounding space (3D vision) and could be implemented for the navigation of robots among obstacles.