基于多尺度空间表征的生物启发目标指引导航模型

发布时间：2018-02-16 09:11

  本文关键词： 类脑导航 空间认知 位置细胞 多尺度表征 Q学习　出处：《电子与信息学报》2017年06期 　论文类型：期刊论文

【摘要】：为实现运行体空间认知和自主导航,借鉴生物导航机理,该文提出基于多尺度空间表征的生物启发目标指引导航模型。首先构建不同尺度位置细胞图编码空间环境,采用高斯模型模拟位置细胞放电率,并将其作为Q学习的状态输入,然后采用模拟退火方法完成行为选择,通过多次探索学习使运行体能够正确规划出一条从起始点到目标点的最短路径。仿真结果表明,该方法用于目标指引导航是可行的,相对于单尺度位置细胞空间认知模型,该方法不但符合多尺度空间表征的生物学依据,而且学习速度更快。在存在障碍物的环境中,能够顺利完成目标指引导航任务,并且当障碍物发生变化时具有较好的适应性。
[Abstract]:In order to realize the spatial cognition and autonomous navigation of moving volume and to learn from the mechanism of biological navigation, this paper proposes a biologically inspired target guidance navigation model based on multi-scale spatial representation. Firstly, the coding spatial environment of cell graph at different scales is constructed. Gao Si model was used to simulate the discharge rate of position cells, which was used as the input of Q-learning state, and then simulated annealing was used to complete the behavior selection. Through multiple explorations, the operator can correctly plan the shortest path from the starting point to the target point. The simulation results show that this method is feasible for target guidance and navigation, and can be compared with the single scale position cell space cognitive model. This method not only accords with the biological basis of multi-scale spatial representation, but also has a faster learning speed. In the environment with obstacles, the target guidance navigation task can be successfully completed, and it has better adaptability when obstacles change.
【作者单位】： 空军工程大学信息与导航学院;西安通信学院;
【基金】：国家自然科学基金(61273048,61473308,61603409)~~
【分类号】：Q42;TP18
，

本文编号：1515165