Robotic Adaptability Through Deep Reinforcement Learning