A Novel Reinforcement Learning Method for Efficient Cross-Training Between Real and Simulated Robots