Fabry-Perot model for Reinforcement Learning