为了提高移动机器人的自学习能力,在基于行为控制结构的基础上设计了智能控制结构,该结构引入了强化学习模块。神经网络具有很好的泛化能力,该文提出了基于神经网络的强化Q学习算法,克服了表格式Q学习算法只能应用到离散的状态中并需要大量存储空间的不足,最后结合智能控制结构应用到移动机器人的避障中。实验结果表明,该方法能够使移动机器人通过自学习实现自主避障。 In order to improve the self-learning ability of mobile robot, an intelligent control structure is designed based on the behavior control structure, which introduces the reinforcement learning module. The neural network has a good generalization ability. In this paper, an improved Q learning algorithm based on neural network is proposed, which overcomes the shortcoming that the tabular Q learning algorithm can only be applied to discrete states and requires a large amount of storage space. Finally, The structure is applied to obstacle avoidance of mobile robot. Experimental results show that the proposed method can make mobile robots realize autonomous obstacle avoidance through self-learning.