While having merits of higher speed and knowledge transferring, the proposed algorithm depends less on aforehand knowledge and can weaken the curse of reward lack in the beginning of learning process.

• 所提出的方法在具备递阶再励学习速度快、易于共享等优点的同时，降低了对先验知识的依赖程度，缓解了学习初期回报值稀少的问题。