Reinforcement learning (RL) - Pump