We use reinforcement learning for the optimal control and stabilization of pendulum and double pendulum systems, adressing the swing-up problem