DC Motor Control based on Integral Reinforcement Learning

The paper presents the control of a DC motor using a machine learning technique known as integral reinforcement learning. The integral reinforcement learning control method belongs to the category of intelligent control systems. The main advantage of the integral reinforcement learning method is that it addresses continuous systems while most reinforcement learning methods are developed for discrete systems. The control system is based on a classic structure in reinforcement learning of critical - actor type. The critic is represented by a neural network that evaluates the efficiency of the actions generated by the actor (the correspondent of the controller in conventional control systems). Critic tuning (neural network training) is done online using the technique known as Temporal Difference Learning. The presented technique is tested and analysed both by simulation and implementation on an experimental platform.