【摘要】 针对激光SLAM中Cartographer算法在IMU数据不可靠时，位姿预测误差较大、扫描匹配时间较长的问题，提出了一种改进的Cartographer算法。首先，在预测先验位姿时，使用位姿数据队列末尾的四个数据来计算移动机器人的线速度，并采用加权融合的方法，对移动机器人的旋转增量进行修正；其次，在使用实时相关性扫描匹配校准先验位姿时，先采用稀疏化策略获取候选位姿旋转得分的分布情况，再对候选位姿得分小于阈值的区域进行剪枝处理，最后对保留下的候选位姿区域计算平移加权得分。仿真表明，在IMU数据不可靠时，改进的Cartographer算法能够有效地改善位姿预测效果，并在保证激光SLAM建图质量的基础上，缩短了实时相关性扫描匹配的时间。更多还原

【Abstract】 In order to solve the problem of large pose prediction error and long scan matching time of the Cartographer algorithm in laser SLAM when IMU data is not reliable, an improved Cartographer algorithm is proposed. First, when predicting the prior pose, the linear velocity of the mobile robot is calculated by using the four data at the end of the data queue, and the rotation increment of the mobile robot is corrected by using the weighted fusion method; Secondly, when using real-time correlation scanning matching to calibrate a priori pose, first use the sparse strategy to obtain the distribution of candidate pose rotation scores, then prune the areas whose candidate pose scores are less than the threshold, and finally calculate the translation weighted scores for the reserved candidate pose areas. The simulation shows that the improved Cartographer algorithm can effectively improve the position and attitude prediction effect when IMU data is not reliable, and shorten the time of real-time correlation scanning and matching on the basis of ensuring the quality of laser SLAM mapping.更多还原