1: $, for ensuring the convergence of Q-learning, the learning rate $