基于具有墙角信息的语义地图改进AMCL重定位算法

针对原始自适应蒙特卡洛定位（Adaptive monte carlo localization，AMCL）算法仅利用激光信息存在的缺陷，提出一种基于激光与视觉融合的语义地图进行全局定位，该语义地图融合基于深度学习的目标检测方法提取环境中的墙角语义；利用建立的包含墙角信息的二维语义栅格地图，结合视觉预定位方法及角点周围语义信息表来提高算法全局初始定位的效率和准确性，使得移动机器人可以在少量先验信息和运动的情况下更迅速地实现定位。提出视觉预定位的方法，改进了粒子权重更新方式，再同步结合AMCL算法与环境地图匹配进行精定位。最后通过搭建的移动机器人在不同场景下进行对比试验，验证了该方法的有效性。

Improved AMCL Relocation Algorithm based on Semantic Map with Corner Information

Aiming at the defects of the original AMCL positioning algorithm using only laser information, a semantic map based on laser and vision fusion is proposed for global positioning. The semantic map is fused with a target detection method based on deep learning to extract the corner semantics in the environment; use The established two-dimensional semantic grid map containing corner information, combined with the visual pre-positioning method and the semantic information table around the corner points to improve the efficiency and accuracy of the algorithm's global initial positioning, so that the mobile robot can be used in the situation of a small amount of prior information and motion To achieve positioning more quickly. The method of visual pre-positioning is proposed, the particle weight update method is improved, and the AMCL algorithm is combined with the environment map to perform precise positioning. Finally, a comparative experiment of the built mobile robot in different scenarios verifies the effectiveness of the method.