Synchronous vs Asynchronous Reinforcement Learning in a Real World Robot