【摘要】 首先介绍了基于表格Ｑ－ｌｅａｒｎｉｎｇ 的方法，然后提出了一种用神经网络实现的Ｑ－ｌｅａｒｎｉｎｇ 方法，利用这种方法实现机器人避碰行为学习，并进行了仿真试验．最后讨论了提高强化学习速度的方法．更多还原

【Abstract】 First, the basic Q-learning based on table is introduced. Then the method that Q-learning is realized by using neural network is presented. The avoidance collision behavior learning of robot is implemented by this method and simulations are carried out. Finally, the method about how to improve the rate of the reinforcement learning is discussed.更多还原