Fruit picking is an important part of agricultural production, and automatic agricultural fruit picking robots can improve work efficiency and work accuracy. An automatic fruit picking robot was developed, which mainly included automatic navigation system, picking system, motion system, control and power system. Among them, the automatic navigation system mainly included two parts: LiDAR navigation and GNSS navigation, which can be used to establish maps and plan the working paths. The picking system carried out fruit recognition through binocular stereo vision, and then sent the results to the actuator (manipulator), which included a sixdegreeoffreedom robot arm and twofinger end. The actuator then grasped the fruit stem and cut the grasping fruit stem to complete the fruit picking. The test results show that the designed and developed robot can complete indoor work through LiDAR navigation. The twofinger end effector that cut and grabbed the fruit stem can successfully finish the fruit picking job. And the developed computer control software can complete image acquisition, robotic arm control, road map creation and other operations. The results of LiDAR navigation accuracy test showed that the absolute error of navigation was less than 3.5cm at 1m/s driving speed, which can meet the demand of greenhouse fruit picking.