Abstract: Simultaneous localization and mapping (SLAM) is mainly used in the field of automatic driving and robot autonomous navigation. Lidar SLAM system is widely used in industry because of its high measurement accuracy and insensitivity to light change. But the SLAM algorithm based on lidar has several problems to deal with, such as: 1) the inaccuracy of localization in less structured information or in changing scenes; 2) the poor ability to correct motion distortion. In this paper, we make two improvements for the above problems: 1) in the case of less structured information, the steps of the original algorithm in dealing with iterative degradation are improved, and a method of combining static threshold and dynamic threshold is proposed for dealing with degradation; 2) in the case of intense exercise, the back-end local map on the basis of 1 is reused to provide ICP with more accurate initial pose and point cloud information. The experimental results show that proposed algorithm has good robustness and localization accuracy in severe motion condition and changing environment, and it also avoids the problem of insufficient incentive in SLAM system.