基于几何特征关联的室内扫描匹配SLAM方法

为实现室内结构化环境中仅依靠激光雷达数据进行实时自定位并创建精确的特征地图,提出了一种基于几何特征关联的室内扫描匹配SLAM方法.几何特征关联与匹配方法的优劣很大程度上影响SLAM的实时性和精度.结合室内结构化环境的特点,提出了一种完备端点定义与提取方法,将直线段特征与完备端点进行关联,优化几何特征扫描匹配过程.此外,姿态角收敛是SLAM进行机器人位姿估计和求解一致性的关键.为确保姿态角准确收敛,采用了基于直线拟合认知的姿态角加权几何平均求解方法.实验证明,提出的SLAM方法得到的定位精度在100mm内,建图精度也较高,能胜任室内SLAM.

Indoor Scan-matching SLAM Method Based on Geometric Features Association

To realize the real-time positioning and precise mapping which only using ladar data in structured indoor environment, an indoor scan-matching SLAM method based on geometric feature association is proposed.The performance of geometric feature association and matching method greatly influence the timeliness and precision of SLAM.Combined with the characteristics of the structured indoor environment, the perfect endpoints definition and extraction method proposed associate points and lines together which optimize the process of geometric features matching.Besides, the convergence of heading is the key to the consistency of pose estimation and calculation of the robot in SLAM method.To ensure the convergence of the heading, an heading weighted geometric average method based on the cognitive of line fitting was proposed.Experiments shows that with the application of the SLAM solution proposed, the accuracy of the positioning is limited in100mm,and the accuracy of mapping is high enough and compete for indoor SLAM