Using a Neurosimulator in the Problem of Maximizing the Output Value of a Dynamic Object

This paper considers the problem of finding quasi-optimal control for an object with an incomplete mathematical model. The main difficulty in practice is obtaining a system of differential equations that describe the dynamics of an object. In the absence of an accurate verified mathematical model of the system, the use of classical methods of optimal control theory is impossible. To solve the problem, it is proposed to use a neurosimulator of the control object and its use in the Chernousko-Krylov numerical iterative method, based on the Pontryagin maximum principle. This approach makes it possible to find quasi-optimal control for "black box" type objects. An example of using a neurosimulator in the problem of maximizing the output phase variable of a dynamic system when setting restrictions on the control signal and control time is given. A comparative analysis of the results obtained for an object specified using differential equations and its neurosimulator was carried out, and the advantages and disadvantages of this approach were identified.