针对移动机器人的导航问题 ,本文提出一种具有连续向量输出的再励学习算法(Reinforcement Learning with Continuous Vector output:CVRL) .CVRL具有层次结构 ,低层是单元动作网络层 ,通过组合动作产生连续向量输出 ;高层是建立在组合单元动作空间上的 Q学习单元 ,实现组合动作的选择 .文中给出了 CVRL导航控制器的具体实现 ,并通过仿真实验验证了其有效性 In order to solve the problem of mobile robot navigation, a Reinforcement Learning with Continuous Vector output (CVRL) is proposed in this paper.CVRL has a hierarchical structure, and the lower layer is a unit action network layer, and a continuous vector output ; The higher level is a Q learning unit which is built on the action space of the combinatorial unit to realize the choice of combinatorial actions.The concrete realization of the CVRL navigation controller is given in this paper and its effectiveness is verified through simulation experiments