Abstract: A four-wheel differential robot was used as experimental platform to obtain environmental information based on single-line liDAR, and the navigation performance of the robot in the center line of the tree and the autonomous line breaking of the robot at the end of the line were investigated. The environmental information was obtained by single laser radar, and the method of distance threshold, the adaptive density clustering algorithm and the improved method of least square algorithm processing laser data were used to obtain the center line of the tree line as robot navigation theory path. The robot and navigation path horizontal offset, heading angle and velocity were selected as input to design the control model of real-time tracking navigation path.The results show that when the robot reaches the end of line, it can automatically sense and execute the turn to the next line. In the inter-row autonomous navigation experiment, when the robot linear velocities are 0.2, 0.4 and 0.6 m/s with the initial lateral offset of robot not more than 0.8 m and the heading angle not more than 15°, the average lateral deviation between the navigation trajectory and the center line of tree line is 3.69 cm, and the maximum lateral deviation is 11 cm. In the autonomous line-breaking test, the robot can accurately turn around to the next line and then continue to navigate between the lines. The proposed orchard autonomous navigation system has good operation stability and high navigation accuracy, which can meet the autonomous driving accuracy requirement of orchard robot.