Research On Visual Servo Control Method For Apple Picking Manipulator

China is a major apple-growing country,which is still dominated by manual picking with low production efficiency.In recent years,countries have committed to developing apple picking robots to reduce labor burden and improve production efficiency.However,the growing environment of apples is complex,and picking robots need to have high-precision recognition and positioning capabilities to ensure successful picking.Most of the current picking robots use open-loop vision control,which is prone to error and leads to picking failure.In addition,the position of the fruit is susceptible to change due to interference from external factors,which can also affect the picking effect.At the same time,the fragility and the vulnerability of the apple skin cause the robot to pick the fruit easily and cause damage to the fruit,which reduces the economic efficiency,and this is also a major problem.In this paper,we design an apple picking robot with embedded multiple sensors and propose a segmented vision servo control strategy for positioning and picking to address the problems that the current apple picking robots will change the position of the fruit when picking,the positioning accuracy is not high,and the apples are easily damaged by grasping.The main research contents of this paper are as follows:(1)An under-driven apple picking manipulator with multiple sensors embedded is designed.A thin-film force sensor is embedded on the finger surface,and the data is filtered and calibrated to provide feedback on the contact force between the finger and the apple,and an incremental PID is used to achieve constant force gripping.A miniature monocular camera and a laser distance measuring sensor were installed in the palm to provide the basic conditions for segmented visual servo control.(2)Target detection and tracking algorithm research was carried out.Target detection and target segmentation were carried out using YOLOv5 s and Open CV,and the combination of deep learning and image processing was proposed to extract target feature information,which effectively improved the accuracy of feature extraction and provided the necessary conditions for segmented visual servo control.KCF(Kernelized Correlation Filters)target tracking algorithm is added to the target detection algorithm to track the target with the highest confidence,which effectively avoids the interference of multiple targets in the camera field of view at the same time to visual servoing.(3)A segmented visual servo control strategy is proposed,and the control strategy is divided into two stages,target alignment and approaching target,to process the video stream and improve the positioning accuracy by using a closed-loop visual control method.The radius of the target-fitting circle is added as feedback in the target alignment stage to effectively shorten the servo path.And the Single Neuron PID algorithm is used to optimize the target alignment stage,which solves the overshoot phenomenon of the target in the fast alignment process,improves the robustness of the target servo control for different distances,and reduces the number of computational iterations.(4)Build an apple picking experiment platform,which mainly includes segmented vision servo control experiment and robotic constant force grasping experiment.The experimental results show that the segmented vision servo control strategy proposed in this paper can perform dynamic target grasping and effectively avoid picking failure caused by external interference,with high positioning accuracy and picking success rate;the designed manipulator can perform constant force grasping and effectively avoid fruit damage.