Reinforcement Learning Approach to Fly Quadcopters with a Faulted Rotor