Active Noise Controller with reinforcement learning

This paper presents a new solution for Active Noise Control problem based on Q-Learning algorithm. This feedback method, needs no information about primary and secondary transfer functions and it is fully robust to subsystem dynamics changes. It is shown through simulation that the proposed method can work properly for a single tone periodic sinusoidal acoustic noise even thought some parameters are changed during the operation. Then, this is extended for multi tones periodic noise and its proper function is satisfied by simulation.