Mobile Robot Localization on a Map with Large Inaccuracy
Mobile Robot Localization on a Map with Large Inaccuracy
复制标题
移动机器人在地图上的定位误差较大
DOI：
10.7210/jrsj.20.425
复制
发表时间：
2002
期刊：
影响因子：
--
通讯作者：
S. Yuta
中科院分区：
文献类型：
--
作者：
M. Tomono;S. Yuta
使用地图进行移动机器人导航的一个主要问题是精确的地图需要大量的建设成本。使用粗略测量的地图的导航框架将是这个问题的解决方案。该论文提出了一种存在描述不准确的拓扑几何图（简称TG图），以及一种在TG图上进行移动机器人定位的方法。 TG 地图是通过定义环境中几何实体之间的相对位姿来构建的。几何实体的模型应该是预定义的，它们之间的相对姿势可能像肉眼测量的那样不准确。这些功能使得地图构建成本很小。机器人位姿以附加在每个实体上的局部坐标系表示，因为基于全局参考系的机器人位姿可能由于地图的不准确而不一致。实体之间相对位姿的误差由概率密度函数表示，并且使用马尔可夫定位的变体来估计机器人位姿，该变体被增强以将地图误差合并到数据融合过程中。仿真和实验表明，该方法能够在TG图上正确估计机器人位姿。
A major problem in mobile robot navigation using a map is that an accurate map needs a lot of building cost. A framework of navigation using a roughly measured map would be a solution of this problem. The paper proposes a Topological-Geometrical map, or TG map for short, which permits inaccurate description, and a method of mobile robot localization on a TG map. A TG map is built by defining the relative poses between geometrical entities in the environment. The models of the geometrical entities are supposed to be predefined, and the relative poses between them can be as inaccurate as those measured by eye. These features allow the map-building cost to be small. The robot pose is represented in a local frame attached on each entity since the robot pose based on a global reference frame might be inconsistent because of the inaccuracy of the map. Errors in relative poses between entities are represented by probability density functions, and the robot pose is estimated using a variant of Markov localization which is augmented so as to incorporate map errors into data fusion process. Simulation and experiments show that the robot pose is estimated correctly on a TG map by the proposed method.