Application of the Reinforcement Learning method for adaptive electric drive control with variable parameters

In this work an artificial neural network was used, which learned using Reinforcement Learning algorithm to control a non-stationary object with a complex mechanical structure that depend on the angular position of the shaft. Critic is presented as a function of control error and control cost, which ensures stability of the system in a long-term performance, without the need to disable the adaptation algorithm. Simulation and experimental results are presented.