A.A. Zhdanov - Dr.Sc. (Phis.-Math.), Professor, JSC Institute of Precision Mechanics and Computer Engineering named by S.A. Lebedev (Moscow). E-mail: a.zhdanov@mail.ru N.B. Preobrazhensky - Ph.D. (Eng.), Senior Research Scientist, JSC Institute of Precision Mechanics and Computer Engineering named after S.A. Lebedev (Moscow). E-mail: nbp@mail.ru Yu.A. Kholopov - Lead Engineer, JSC Institute of Precision Mechanics and Computer Engineering named after S.A. Lebedev (Moscow). E-mail: hol_it_m@mail.ru I.V. Stepanian - Dr.Sc. (Biol.), Ph.D. (Eng.), Leading Research Scientist, Laboratory of biodynamics, Institute for Machine Science named by A.A. Blagonravov of the RAS, Leading Research Scientist, Moscow State Сonservatory named by P.I. Chaikovsky (Moscow). E-mail: neurocomp.pro@gmail.com Nguyen Huu Chung - M.Sc., Moscow Institute of Physics and Technology

Nowadays, most of automatic control systems are based on rigorous mathematical models. However, this approach has some limits when being applied in practice, since it-s extremely difficult to form rigorous mathematical models for some object classes. When the complexity of the controlling tasks increases, building their accurate mathematical models seems to be impossible. Since the mid-twentieth century, it began to spread the so-called \"intelligent\" approach which was used to create controlling systems. In this case, instead of using mathematical models of the object, the intelligent approach use and knowledge of object. This method of constructing control system eliminates the step of creating an accurate mathematical model of the object. \"Autonomous Adaptive Control\" (AAC) method allows us to create automatically models of object that is used to control the object in accordance with user-defined objectives and criteria without its mathematical description. Separate subsystem of AAC system can be implemented with various appropriate methods of pattern recognition, knowledge representation and so on. However, the most interesting and appropriate manner of constructing AAC system is using a neural-like network method corresponding to all the visible properties of biological control systems. Because available neuron models are not appropriate to AAC, we created specific neutron models which are described in [Zhdanov A.A. Self-contained artificial intelligence. - M.: Binom. Laboratory Knowledge 2008]. The most distinguishing feature of our model of a neuron is that biological neurons is considered as a separate self-learning recognition system. Our further desire is to build the AAC neuron system not in software form, but in the form of ASIC with a high degree of parallelism as possible - down to a single neuron. The first steps of implementing the task are going to be presented in this paper. This work was financially supported by RHSF in grant № 15-03-00519а «Post-non-classic paradigm of artificial intellect».

 

1. ZHdanov A.A. Avtonomnyjj iskusstvennyjj intellekt. M.: BINOM. Laboratorija znanijj. 2008.
2. ZHdanov A.A. Formalnaja model nejjrona i nejjroseti v metodologii avtonomnogo adaptivnogo upravlenija // Sbornik «Voprosy kibernetiki». Nauchnyjj sovet po kompleksnojj probleme «Kibernetika» RAN. 1997. Vyp. 3. C. 258-274.
3. Neurox Intellect Labs // http://neurox.intellect-labs.com
4. Modelsim // http://www.altera.com/products/design-software/model---simulation/modelsim-altera-software.html.
5. Ieruzalimski R. Programmirovanie na jazyke LUA. M.: DMK Press. 2014. 382 s.