It can be seen from the simulation results that with the MOVEMENT of AGV in a complex environment, the path prediction error and landmark prediction error become larger and larger, because of the accumulation of errors.Ekf-slam takes about 26 minutes, and the maximum deviation occurs in the Y-axis direction, which is about 0.6 meters.Ukf-slam takes about 30 minutes, and the maximum deviation occurs in the X-axis direction, which is about 0.12m.It can be seen from FIG. 5.9 and 5.10 that the error of UKE-SLAM is far less than that of EKF-SLAM. In the complex map environment, UKF-SLAM has a better adaptability than EKF-SLAM.The path deviation of the two algorithms mainly occurs in the path point corner. With the AGV motion, the error accumulates gradually, and the prediction of the path punctuation produces more and more errors, thus affecting the prediction of the motion path.