A constrained SLAM approach to robust and accurate localisation of autonomous ground vehicles
A constrained SLAM approach to robust and accurate localisation of autonomous ground vehicles
复制标题
一种用于自主地面车辆稳健且准确定位的约束 SLAM 方法
DOI：
10.1016/j.robot.2007.02.004
复制
发表时间：
2007
期刊：
影响因子：
--
通讯作者：
J. Guzman
中科院分区：
文献类型：
--
作者：
K. Lee;S. Wijesoma;J. Guzman
为了使驾驶辅助系统、智能车辆和自主机器人在复杂环境中可行，必须具有可靠且强大的定位功能。由于这种复杂环境（包括主题公园、大学校园、郊区、工业区等）的大的可变性和不确定性，难以依赖特定的方法或传感器数据集来正确且鲁棒地估计机器人路径/姿态。解决本地化问题的关键是优化使用和融合移动的平台可用的所有有用信息源。在设想的环境中，拥有道路网的近似数字地图并不罕见。在本文中，除了典型的感官信息所提供的extereoceptive和本体感受传感器，它示出了如何先验近似的知识，可在一个同时定位和地图建设（SLAM）的框架内，系统地融合在一个路线图的形式，以获得更准确和更强大的定位结果。通过以先验地图信息的形式引入约束，这种SLAM的重新表述不仅使问题在理论上在可观测性的意义上更正确，而且使系统可行且有效，产生更准确的结果。在实际环境中获得的结果来验证索赔。
For driving assistance systems, intelligent vehicles and autonomous robots to be viable in complex environments, it is necessary to have a reliable and robust localisation function. Due to the large variability and uncertainty of such complex environments, which include theme parks, university campuses, suburbs, industrial estates and the like, it is difficult to rely on a specific method or set of sensor data to correctly and robustly estimate the robot path/pose. The key to solving the localisation problem is to optimally use and fuse all useful sources of information available to the mobile platform. For the envisaged environment, it is not unusual to have approximate digital maps of the road network. In this paper, in addition to the typical sensory information provided by extereoceptive and proprioceptive sensors, it is shown how a priori approximate knowledge available in the form of a road map can be systematically fused within a Simultaneous Localisation and Map Building (SLAM) framework to obtain more accurate and robust localisation results. This reformulation of SLAM through the introduction of constraints in the form of a priori map information not only makes the problem theoretically more correct in the sense of observability but also makes the system viable and effective, yielding more accurate results. The results obtained in an actual environment are presented to validate the claims.