Universal Robotic Picking End-Effectors for Various Fruits (www.mdpi.com)

This review maps the state of research toward universal robotic fruit-picking end-effectors, synthesizing designs, materials, sensing and control trends aimed at handling wide varietal diversity and unstructured canopy environments. It classifies end-effectors by picking pattern (grasp-and-pull, -twist, -bend, -cut and combined actions) and contrasts rigid grippers, soft designs (soft pneumatic actuators, SPAs; soft air chambers) and Fin Ray Effect (FRE) structures. Key advances include integration of tactile sensing for closed‑loop force feedback, field-ready materials (ME3P, MREs, LCEs), and learning-enabled control (deep reinforcement learning) paired with vision/spectroscopy pipelines (RGB‑D, YOLO/SOLOv2 variants, hyperspectral) for robust fruit detection and picking-point decisions. Significant technical challenges and research directions are highlighted: creating anthropomorphic master–slave hands with dimensionality‑reduction mappings from human kinematics to simplified robotic DOFs; combined palm–finger kinematic models to enable adaptive envelopment; and rigid–flexible coupling designs that trade compliance for stiffness, load capacity and actuation speed. The review suggests optimizing anisotropic rigid inserts and module layouts, using kinematic and interaction‑force models plus FEA to inform structure and control, and bringing perception, tactile feedback and DRL together for sim‑to‑real, damage‑minimizing harvest. For AI/ML practitioners, the paper underlines opportunities in perception, force-aware policy learning and data-driven kinematic/force modeling to close the gap to deployable, high-TRL harvesting systems.