V.V. Kozhevnikov1, V.V. Prikhodko1, V.V. Svetukhin1, A.V. Zhukov1, A.N. Fomin1, M.Yu. Leontyev1, D.Ya. Vostretsov1, A.A. Sobolev1, V.I. Skrebtsov1, V.E. Kiryukhin1, V.V. Levschanov1, D.S. Lavygin1, E.M. Chavkin1, E.R. Mingachev1, R.G. Bildanov1, S.V. Pavlov2, V.N. Kovalnogov1

1 S.P. Kapitsa Research Institute of Technology (Technological Research Institute) of Ulyanovsk State University, Ulyanovsk, Russia
2 Sosny Research and Development Company, Dimitrovgrad, Ulyanovsk region, Russia

Received 29 June, 2018; accepted in revised form 17 July, 2018

Principal directions of developing the methods for designing intelligent systems of robot control assume technologies based on the use of artificial neural networks. The neural networks, where the model of a neuron was developed as the simplest processor element, performing the computation of the transfer function of a scalar product of an input data vector and a weight vector, can give interesting results regarding generation of dependencies and forecasting. However, their obvious drawback is the lack of an explicit algorithm of action. Memorization of information in the learning process occurs implicitly as a result of selection of the weight coefficients of the neural network, therefore the problem of cognition (the formation of new knowledge) on the basis of those obtained earlier in the learning process seems difficult to resolve. A positive solution to this problem will open the way to the creation of the full-fledged artificial mind. From this point of view the promising area is where the mathematical model of the neural networks is built on the basis of mathematical logic. The intelligent control system in this case is a software and hardware complex, where the mathematical model of the neural network identifies the control system as an intellectual one.

c 2018 European Society of Computational Methods in Sciences and Engineering


Scroll to Top