Abstract: This paper designs a control system of apple picking robot based on crawler motion. The working principle of the system is as follows: firstly, the apple image is collected by the industrial camera installed on the manipulator and sent to the general control system. The general control system uses the deep learning yolov3 algorithm to identify the target Apple and detect its position information; Then, the distance sensor installed on the clamping claw collects the distance information between the apple and the clamping claw, performs pose calculation, and sends the result to the manipulator motion control card, which sends action instructions to make the manipulator reach the target picking position, and the camera head and infrared sensor at the end of the clamping claw will assist in confirming that the apple falls into the clamping claw; At this time, the end execution control module sends the grasping command to control the operation of the stepping motor to make the clamping claw grasp the apple, and feed back the closing degree of the clamping claw through the pressure sensor to avoid scratching the apple; Finally, the end execution control module drives the slicing motor to cut branches and pick apples. The experimental results show that the probability of accurately picking apples by this method is 86.67%, which can basically meet the needs of apple picking.