Reinforcement Learning target performance estimation on K2D9 500-5-late