Fresh-market citrus must be detached with a short residual stem to prevent puncture damage to adjacent fruit during handling and transport. This paper presents a passive stem-centering end-effector that replaces the active two-axis positioning stage of our previous standalone harvester, and its integration with vision-based stem detection and a collaborative manipulator on a mobile platform built from commercially available components. The platform integrates a 6-degree-of-freedom (DoF) collaborative manipulator and an RGB-D vision system based on YOLOv8 instance segmentation, and the end-effector comprises a compliant three-finger gripper, commercial citrus clippers with concave blades, and a V-shaped stem-centering guide. Coordinated manipulator roll and pitch motions guide the stem into the guide toward the base of the blades, which reduces the end-effector mass from 1.5 kg to 0.5 kg and the number of actuators from four to two. In supervised orchard trials on 100 navel oranges, all target fruit were detached; 92% met a residual stem length of less than 10 mm and 52% met the stricter 2 mm requirement, with a mean residual stem length of 3.10 ± 3.34 mm (median 1.81 mm). Fruit without adverse stem or canopy conditions achieved 1.01 ± 0.56 mm, close to the result of the previous active design, whereas longer residual stems were associated with non-uniform stems, extreme stem angles, branching stems, or obstructions. The complete harvesting cycle required 38.6–55.1 s per fruit, depending on manipulator speed. The results indicate that passive stem-centering can provide controlled residual stem length with simpler end-effector hardware.
No takes yet. Share an insight, caveat, or question.
Chinnock et al. (2026) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: