The localization drift caused by the degeneracy problem and dynamic environments is a major challenge in LiDAR SLAM. This paper proposes a robust 2D LiDAR SLAM framework for industrial logistic scenarios, combining multiple landmark constraints to address these two issues. Three types of landmarks are integrated into the existing method, namely, retro-reflective column, planar reflector, and Data Matrix (DM) code. Customized methods are devised to extract the information of the first two types of landmarks through the intensity value of LiDAR measurements, while the DM codes are read by the on-board camera. The special constraint equations corresponding to these landmarks are proposed and incorporated into the existing backend framework. The experimental results show that both mapping and localization accuracy are significantly improved in degeneration scenarios and dynamic environments.