Aiming at the problem of motion estimation of moving objects in dynamic scenes, the simultaneous localization and mapping (SLAM) method is analyzed. The existing dynamic SLAM algorithms can be divided into 2 categories: the geometry-based method and the deep learning-based method. This paper focuses on a geometry-based method, the dynamic SLAM method based on the correlation of point clouds, summarizes the challenges faced by the current SLAM technology, and looks forward to the development potential and direction of the dynamic SLAM for future warfare. The results show that the research can promote the application of SLAM technology in robot navigation.