ANN Training Method with a Small Number of Examples Used for Robots Control

This paper presents a method for obtaining a neural model used in industrial robots control. The method refers to the forming of a small number of examples used in the training of a neural network that lead to the creation of a suitable model. This paper constitutes a development of the work [2] in order to increase the opportunities for its application in various fields. The description of the method is generally done, without relying on a specific application in the domain of industrial robots. The testing and the validation of the shown method were completed using the example of a system in which the relationship between inputs and outputs is described by means of mathematical functions. The set of learning examples, generated through the proposed method, served to the ANN training by a cross-validation technique, in case of these functions. The evaluation of the proposed method has been done by analysing the results obtained by applying it compared to those obtained with a known method, namely the uniform generation of training examples. The use of the method in the field of industrial robots’ control was illustrated by a concrete application in the case of a robot with 6 degrees of freedom.