Scientific journal
Scientific Review. Technical science
ISSN 2500-0799
ПИ №ФС77-57440

Personal portfolio

NEURO-CONTROLLEROF QUADCOPTER CONTROL SYSTEM

Evgenov A.A. 1
1 FSBEI HPO «South-RussiaStatePolitechnicalUniversity (NPI) n. a. М. I. Platov»
In this paper is proposed a problem solving of complex dynamic object control – quadcopter by neuro-controller. Initial values of parameters are definedanalytically and neural network implements the search of controller values whereby the error ofevaluate function, that demonstrates error in following some reference trajectory, is going to zero.This method is suboptimal as no way to divine: is obtained parameters optimal or it’s a local optimum.There is only a possibility to increase chance to get optimal coefficients by dilation of neural network training set. The analysis results of modeling show that neuro-controller is more effective than PID-controller. Obtained figures of transient response of pitch, yaw and roll angles and angular velocity of neuro-controller have smaller amplitude damped oscillations and faster response time then customary PID-controller. Neuro-controller is able to adapt to changes of external influences.