AI-driven adaptive grasping and precise detaching robot for efficient citrus harvesting

Abstract

Many agricultural tasks in open field such as fruit harvesting must be conducted during specific periods and are labor-intensive, making it difficult to provide workers on time. Automating these tasks is challenging due to unstructured orchard workspaces, variations in environmental conditions such as lighting, dense plant growth with many occlusions, varying climates, and the need for sophisticated manipulation of soft fruit objects. In this paper, we present an AI-driven citrus harvesting robot system capable of adaptive grasping and precise detaching. The proposed robot features an eye-in-hand manipulator with an adaptive grasping and precise detaching end-effector. Its perception system detects the 6D pose of fruit instances in real-time and generates appropriate harvesting motions to cut the peduncle of the target fruit, thereby minimizing the remaining peduncle. We propose an integrated control system for the end-effector and manipulator to perform autonomous harvesting tasks. We evaluated the performance of the proposed harvesting robotic system in two experimental conditions: mock-up citrus and real citrus orchard. The harvesting robotic system demonstrates high success rates and fast harvesting speeds, making it suitable for practical orchard applications.