Abstract: Factors such as sensors error and inaccurate robot positioning can cause errors and distortions during mobile robot mapping. To address this problem, a map optimization method is proposed, which improves the traditional Gmapping algorithm. Firstly, an improved Iterative Closest Point (ICP) algorithm is introduced for the scanning data of lidar to match the laser point cloud between frames, which can improve the registration effect of laser point cloud. Secondly, the particle resampling step of the traditional Gmapping algorithm is improved by dividing particles into three categories based on weight, and the diversity of the particles is increased by classification resampling. Finally, a mobile robot equipped with a 2D lidar was used in the experimental environment, and the SLAM map optimization algorithm experiments were conducted by the Robot Operating System (ROS). Experimental results demonstrate that the optimization method can improve the mapping accuracy significantly and resolve map distortion issues effectively.