Design and experimental investigation of the grasping system of an agricultural soft manipulator based on FMDS-YOLOv8

In response to the need for non-destructive sorting and grasping of fruits and vegetables with diverse sizes and shapes, this study presents a novel design for an agricultural manipulator grasping system (MGS). The system includes a variable-structure soft manipulator equipped with three independently rotatable and distance-adjustable soft actuators. The manipulator can grasp objects with a diameter of ≤140 mm in the center grasping configuration and ≤105 mm in the parallel grasping configuration. An improved FMDS-YOLOv8 vision recognition algorithm was used to detect the type, contour and positional coordinates of the target fruit. A MATLAB-based program was developed to extract the contours of the target fruit and calculate the visualization of the optimal attitude of the soft manipulator. This program facilitated autonomous structural adjustments and precise control during grasping operations. The variable-structure soft MGS was evaluated based on the performance of each component. The experimental results showed a grasping success rate of 95.83%, a grasping damage rate of 4.17%, and a grasping time of about 6.36 s under multi-objective conditions. This verifies the effectiveness and adaptability of the MGS. By adjusting the drive pressure and servo angle, the MGS can grasp fruit and vegetables of different sizes and shapes within its working range, while minimizing damage during the grasping process.