This paper presents a method for mobile robots to recognize and dock shelves based on 2D LiDAR, which can autonomously identify shelves and perform high-precision docking. A geometric feature-based shelf leg recognition and multi-shelf model screening method is designed, which can quickly and accurately autonomously identify shelves. A fitting method for the center of shelf legs and shelf centers based on nonlinear least squares was proposed, and a real-time autonomous docking system was designed, which can autonomously distinguish shelf size types and perform precise docking. The experiment shows that the proposed method can achieve a docking accuracy of ± 1 cm.