Adaptive Policy Learning to Additional Tasks