A method of constructing neuro-fuzzy controller based on adaptive algorithm of self-organizing network to control the angle of heel of the unmanned aerial vehicle

Abstract

The article discusses the method of designing, training and use of the neuro — fuzzy controller to control the behavior of the angle of heel of the unmanned aerial vehicle. The regulator is constructed in the form of a fuzzy neural network, the inputs of which are fuzzy linguistic variables such as the deviation of the angle of heel from the nominal impact, speed and acceleration, while output is a clear option, that is the control action exerted on the object of regulation. To train the network an adaptive algorithm of self-organizing fuzzy network is used, which allows building the architecture of the fuzzy neural network based on the source data and the Gaussian membership functions. The technique of designing the training and testing samples, based on knowledge of the desired behavior of the object under different nominal impacts, is proposed. The results of experimental research on selection of parameters of membership functions and using designed controller are given.